Interaction between Human CD2 and CD58 Involves the Major ~ Sheet Surface of each of Their Respective Adhesion Domains By Antonio R. N. Arulanandam,*~ Alexander Kister,*~ Malcolm J. McGregor, 82 Daniel F. Wyss,$ Gerhard Wagner,~ and Ellis L. Reinherz*II From the *Laboratory of Immunobiology, Dana-Farber Cancer Institute and the Departments of IPathology, ~Biological Chemistry and Molecular Pharmacology, and IlMedicine, Harvard Medical School, Boston, Massachusetts 02115; and 82 Inc., Cambridge, Massachusetts 02139 he interaction of human CD2 on T lymphocytes with CD58 on the surface of antigen-presenting, epithelial, T endothelial, and target cells of various types is critical for T lymphocytes to mediate their regulatory and effector functions (1-3). The adhesion domains of CD2 and CD58 facilitate the antigen recognition process by stabilizing cell-cell contact (4, 5). In addition, ligation of CD2 by CD58 or antiCD2 mAbs provides signals via the CD2 cytoplasmic tail, thereby lowering the activation threshold for TCR triggering (6-11). CD2 also facilitates cytotoxic function of NK cells through similar mechanisms (12-14). The membrane distal NH2-terminal domains of human CD2 and human CD58 bind to one another with a micromolar affinity (15-17). This modest, monomeric binding activity is amplified by multimeric interaction at the cell-cell interface resulting from rapid redistribution of CD2 to the intercellular junction (5, 18). Human CD2 also interacts with sheep CD58, providing the molecular basis for the fortuitous observation that human T cells and sheep erythrocytes form aggregates, or "rosettes" in vitro (19-23). This rosetting 1861 phenomenon was used to enumerate T lymphocytes in humans long before the discovery of CD2 (20-23). Human CD2 interacts with a second structurally related human ligand, CD48 (24). However, the affinity of the CD2-CD48 interaction is 2 orders of magnitude weaker (10 -4 M) than that of the CD2-CD58 interaction and hence, is unable to support cell adhesion in an in vitro assay (24). In the murine system, in contrast, CD48 is the only known ligand for CD2 (25). Thus, it appears that the ligands of CD2 have diverged during the evolution of humans and rodents (24). Molecular cloning of human CD2 and CD58 indicated that their gene products are r homologous at the amino acid level and suggested their structure to be Ig-like (26-30). Nuclear magnetic resonance (NMR) 1 spectroscopy and x-ray crystallographic studies confirm that both rat and human CD2 adhesion domains adopt the predicted Ig fold (31-34). Human 1 Abbreviationsused in thispaper: aa, aminoacid;NMR, nuclearmagnetic resonance. j. Exp. Med. 9 The RockefellerUniversityPress 90022-1007/94/11/1861/11 $2.00 Volume180 November1994 1861-1871 Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 Summary The CD58 binding site on human CD2 was recently shown by nuclear magnetic resonance structural data in conjunction with site-directed mutagenesis to be a highly charged surface area covering "~770A2 on the major AGFCC'C" face of the CD2 immunoglobulin-like (Ig-like) NH2-terminal domain. Here we have identified the other binding surface of the CD2-CD58 adhesion pair by mutating charged residues shared among CD2 ligands (human CD58, sheep CD58, and human CD48) that are predicted to be solvent exposed on a molecular model of the Ig-like adhesion domain of human CD58. This site includes B strand residues along the C strand (E25, K29, and K30), in the middle of the C' strand (E37) and in the G strand (K87). In addition, several residues on the CC' loop (K32, D33, and K34) form this site. Thus, the interaction between CD2 and CD58 involves the major B sheet surface of each adhesion domain. Possible docking orientations for the CD2-CD58 molecular complex are offered. Strict conservation of human and sheep CD58 residues within the involved C and C' strands and CC' loop suggests that this region is particularly important for stable formation of the CD2-CD58 complex. The analysis of this complex offers molecular insight into the nature of a receptor-ligand pair involving two Ig family members. Materials and Methods Antibodies. TS2/9 mAb was providedby Dr. T. Springer (Center for Blood Research, Boston, MA). All other anti-CD58 mAbs used in this study were generouslyprovidedby Dr. StefanMeuer (German Cancer Research Center, Heidelberg, Germany) (45). The CD58 heteroantisera, Rb202, was provided by Dr. Laurlee Osborn (Biogen Inc., Cambridge, MA) (17). Site-directedMutagenesis. A 0.8-kb CD58 cDNA fragment was excised from CDM8 (28) with XbaI and subcloned into the XbaI site of M13mp18 in order to generate site-directed mutants. Mutagenesis and subcloning of mutant CD58 cDNA molecules into the CDM8 expression vector was conducted as previously described (43). The entire adhesion domain of each variant was sequenced to exclude second-site mutations. Transfection, Binding Analysis, and ImmunofluorescenceAssays. CHO cells were cultured and transfected with CD58 cDNAs as previously described (24). Stable CHO cell lines expressing mutant CD58 at copy numbers comparable or higher than wtCD58transfected CHO cells were derived by fluorescence-activatedcell sorting of the parental transfectants. For this purpose, we used a rabbit heteroantisera specificfor the extracellular segment of CD58, termed RB202 (1:100 dilution) followed by staining with FITCconjugated goat anti-rabbit Ig (Tago, Inc., Burlingame, CA). For binding studies, CHO cells expressing mutant or wtCD58 mole1862 cules were plated in triplicates at 5 • 104 cells/well in 24-well plates and analyzed for CD2 binding using SlCr-labeled Jurkat cells as previously described (24). For experiments aimed at identifying anti-CD58 mAb binding residues on human CD58, nonsaturating concentrations of mAbs (yielding between 50 and 90% of maximum fluorescence)on wtCD58-transfected CHO cellswhen developedusing a goat anti-mouse Ig labeledwith FITC (BioWhittaker, Inc., Walkersville, MD) were used to optimize sensitivity. Cell-bound immunofluorescencewas determined using a FACScan| (Becton Dickinson & Co., Mountain View, CA) as described (see legend to Table 1). MolecularModelingof the Human CD58 AdhesionDomain. Two three-dimensional molecular models of the adhesion domain of human CD58 were developed independently, based on homology with the known three-dimensional NMR structure of the human CD2 adhesion domain (32, 33). The first model was based on a structural alignment of human CD2, rat CD2, human CD4, human CD8, and REI Ig molecules (31-33, 46-48). This produced a sequence alignment to which the CD58 sequence was aligned. The coordinates of the main chain secondary structure regions were taken from human CD2 (average of 18 newly refined NMR structures, energy minimized). The conformations of the loops were modeled on the structure whose sequence in the loop most closely resembles that of CD58. For example, the conformation of the CC' loop was modeled on the corresponding loop in rat CD2 (involving a two-residue deletion in human CD2). The C'C" loop was formed by deleting E52 in human CD2, which is a bulge at the beginning of the C" strand, and preserving the conformation of the turn. The C"D loop was modeled on that of CD4 in order to form a salt bridge between residues R52 and D71. This salt bridge is conserved in most Ig domains, but is not present in CD2. For the FG loop, the alignment of the F and G strands means that if the strands form a r this puts proline 80 in the first position of a four-residue ~-turn; this is unusual and introduces a conformational distortion. There is no similar loop in the above structures but one was found in the FG loop of a mouse Ig H chain constant domain (Brookhaven structure 1FAI) and this was incorporated into the model. The side chain conformations were chosen using the above structures as a guide and ensuring that no severe steric clashes were created. In the second model, a secondary structural alignment based on human CD2 was used and the conformation of loops chosen to provide best conformity with the experimentally derived mAb and CD2 binding results. Both models were energy minimizedusing the Amber force field in the Discover program from Biosyn Technologies (San Diego, CA). The two CD58 models differ slightly in the alignment of the A and C" strands, and in the EF loop (data not shown). However, they are in agreement as to the alignment of the important F, G, C, and C' strands where most of the mutations are located. Results and Discussion SequenceAlignment of the Adhesion Domains of CD58, CD48, and CD2. The N M R studies of rat and human CD2 adhesion domains show that each adopts an Ig fold (31-33). These solution structures were confirmed and extended by x-ray crystallography analyses of the entire rat CD2 ectodomain (34). The second domain was found to be Ig-like as well and connected to the membrane-distal first domain by a flexible hinge region. Given the weak but nevertheless significant homology between the extracellular segments of human CD2 CD2-CD58Molecular Interaction Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 CD48 shares somewhat higher homology (,030%) with both human CD2 and CD58, implying that human CD58 may have evolved from CD48 (24, 35, 36). The human CD2 and CD58 genes are located on chromosome lp and the CD48 gene is on lq (36, 37). These genes are tightly linked with the ATP1A genes on chromosome 1 which have evolved by gene duplication (36-38). By inference, it is likely that duplication of a primordial gene also gave rise to the structurally related CD2, CD58, and CD48 genes (36, 38). In this regard, the human genomic organization of CD2 and CD58 is similar except for an additional exon in human CD58 which contains the alternative splice sites for the phosphatidyl-inositol (PI), linked and transmembrane forms of CD58 (39, and Wallich, R., personal communication). However, the tissue distributions of CD58, CD48, and CD2 markedly differ from one another. CD58 is ubiquitously expressed in many diverse cell types, whereas CD48 expression is restricted to leukocytes and human CD2 is limited to T and NK cells (40-42). Site-directed mutagenesis studies based on the N M R structure of the adhesion domain of human CD2 defines the CD58 binding surface as a highly charged surface area restricted to the AGFCC'C" face of its Ig-like domain (43). An independent mutagenesis study predicated on a homology model of the human CD2 adhesion domain yields similar conclusions (44). Given these findings, it seems likely that a charged surface on the NH2-terminal domain of human CD58 might bind to CD2. Based on this speculation and the known sequences of human CD2 ligands (human CD58, human CD48, and sheep CD58), we selected charged residues on human CD58 that are conserved among the CD2 ligands for sitedirected mutagenesis. The ability of mutant CD58 molecules to bind CD2 was then analyzed and the binding site for human CD2 on CD58 was mapped. In addition, the predicted structural features shared in common by the other CD2 ligands permits us to infer how they might bind to CD2 as well. 2930 3 ~ 25 9 37 r162 9 CD58 ( hum ) ~~z/_~s CD58 ( sheep ) ~ ~ - -[.~TT~ -- ~ ~...... CD48 ( hum ) CD2 (hum) 7 12 16 A 39 CD58 ( hum ) CD58 ( sheep ) CD48 ( hum ) CD2 (hum) 44 21 32 n 50 52 37 45 C 56 47 C' 7273 76 78 84 87 ENSE-F Y S .... ITDT~ ~TSGL ....... T I ....VKKS ~SSKY ~ P SKVKV~N ~ N SI RVLK-~ EQ~W~IKI V ~IKI W RK@TF~ - - ~ D T Y ~ F K - ~ ] K ~ K T ~ I ~ V SIYDT~VL~K I~KI Q 53.55 60 63 C" D 67 71 80 E 86 Arulanandam et al. 103 G Generation and Expression of Human CD58 Mutant Molecules. Residues conserved among at least severalof the CD2 ligands seem the most obvious candidates to be implicated in the CD2 contact site. With these criteria in mind, we selected 15 conserved residues in domain I which are charged and surface exposed (Fig. 1) for alanine scanning-based sitedirected mutagenesis studies. In addition, three charged but nonconserved residues each localized in the predicted C'C" loop, C" and G strands were also mutagenized. As shown in the schematic representation of the human CD58 domain (Fig. 2), 10 of the 18 residues mutated map within/3 strands: 9 TS2/9 9 IA3 9 IH10 9 IA2 93 Figure 2. Location of CD58 point mutations and anti-CD58 mAb binding sites on a molecular model of the adhesion domain of human CD58. Ribbon diagram representation of a "MOLSCRIPT" drawing (56, 57) showing the predicted AGFCC'C" B sheet in the front (light) and the BED sheet in the back (dark) with residue numbers indicating the beginning and end of each/3 strand. Residues analyzed by site-directed mutagenesis to identify binding sites for CD2 and various anti-CD58 mAbs are shown circled and their positions on the CD58 model, except for P80, are based on both sequence alignment and CD2/3 strand assignments (Fig. 1, and 32, 33). Point mutants that affect anti-CD58 mAb binding are indicated with symbols: TS2/9 (O), 1A3 (11), 1H10 (A), and 1A2 (V). Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 and human CD58, it is likely that CD58 also consists of two Ig-like domains (28). The NH2-terminal domain (amino acid [aa]1 1-93) of CD58 has been shown to contain the CD2 binding region (17). Fig. 1 offers an alignment between human CD2 and human CD58 adhesion domains and indicates the position of the CD2/3 strands as defined by NMR. The conservation of multiple key amino acid residues found in Ig domains implies that the CD58 adhesion domain adopts an Ig-fold. An invariant tryptophan (W28) in the predicted C strand is present at the same position in all known Ig supergene family members. In addition, other CD58 residues including V13, F15, V35, R52, L62, I64, D71, Y75, L90, and V92 are conserved among Ig family members (Fig. I and data not shown). R52 at the end of the predicted C"D loop and D71 in the predicted EF turn presumably form a salt bridge that is characteristic of most Ig molecules. This salt bridge is absent in CD2. In addition, the human CD58 adhesion domain, like the CD2 adhesion domain, also differs from conventional Ig domains in lacking two cysteine residues which form an intradomain disulfide bond in almost all Ig domains (28-30). Valine (V17) at the end of the predicted CD58 B strand and methionine (M77) at the predicted F strand presumably function in lieu of the cysteine residues to mediate hydrophobic contacts (Fig. 1). In this regard, replacement of the corresponding hydrophobic residues on rat CD2 in the B and F strands by cysteinesthrough site-directed mutagenesis showed that a disulfide bond forms at this position (49). Hydrophobic residue (V13) is located near the start of the B strand and Y75 is located two residues NH2-terminal to M77, near the beginning of the F strand. These hydrophobic residues are hallmarks of Ig domains and form part of the hydrophobic core of the Ig domain (30). Fig. 1 also depicts the sequence similarity between the NH2-terminal domain of sheep CD58 and human CD2 as well as human CD48 and human CD2. Human CD48 consists of two Ig-like domains as well (35). As noted, there is considerable homology among the NH2terminal domains of the three human CD2 ligands (human CD58, human CD48, and sheepCD58): 55% between human CD58 and sheep CD58 and 30% between human CD58 and human CD48 (Fig. 1). These ligands all possess key conserved Ig domain residues. 1863 94 F Figure 1. Alignment of amino acid sequences in the adhesion domains of CD58, CD48, and CD2. The amino acid sequences were aligned as described in Materials and Methods. The positions for the/~ strand and loop residues in the human CD2 adhesion domain (32, 33) are shown as defined by NMR analysis. Residues that are conserved between all sequences are boxed and shaded. CD58 residues that are involved in binding CD2 are shown in either closed circles (strong effects) or half-closed circle (partial effect). Analyzed residues not involved in binding CD2 are shown in open circles (no effect). Sheep CD58 sequence data are available from GSDB/DDBJ/EMBL/NCBI DNA under accession number D28584. two residues in the G strand (D84 and K87), two residues in the F strand (E76 and E78), three residues in the C strand (E25, K29, and K30), one residue in the C' strand (E37), one residue in the C" strand (R44), and one residue in the D strand (D56). The other eight residues tested map within the predicted loops: three residues in the CC' loop (K32, D33, and K34), one residue in the C'C" loop (E39), two residues in the C"D loop (K50 and R52), and two residues in the EF loop (E72 and D73). The wtCD58 molecule or individual mutants were transfected into CHO cells and stable lines expressing these molecules were screened by indirect immunofluorescence using a rabbit anti-CD58 heteroantiserum (Rb202). C H O lines expressing CD58 molecules at levels comparable or higher than wtCD58 were selected by immunofluorescence and cell sorting. In this way, we could exo dude that loss of CD2 binding activity was related to a low CD58 copy number on the C H O cells. Specific Cell-based Adhesion Assay to Analyze the Interaction of Human CD2 with CD58 Mutants. The interaction between human CD2 and CD58 was studied in a cell-cell adhesion assay using CD2 + Jurkat cells and C H O cells transfected with CD58 cDNA encoding either wtCD58 or one of the mutant molecules. Previous studies using this assay showed that the binding of Jurkat cells to CD58 + C H O cells is strictly dependent on the CD2-CD58 interaction (24). The photomicrographs in Fig. 3 compare the results of a repre1864 sentative binding experiment between Jurkat and the wtCD58 C H O transfectant on the one hand (Fig. 3 A) and between Jurkat and a CD58 point mutant (K34A) which completely disrupts binding on the other hand (Fig. 3 B). Note the readily apparent aggregates between the round, relatively small Jurkat cells and the large, fusiform CHO cells transfected with wtCD58. In contrast, virtually no aggregates are formed between Jurkat cells and the K34A CD58 transfectant. This result is noteworthy particularly because the level of CD58 surface expression on K34A C H O transfected cells is more than threefold that on the wtCD58 CHO transfectant (Fig. 3 insets, and Table 1). CD58 Point Mutants. The results of experiments with all CD58 mutants expressed in CHO cells are summarized in Table 1. The percentage of binding for individual CD58 mutants was calculated relative to wtCD58 using a previously described semi-quantitative CHO binding assay with 51Crlabeled Jurkat cells (24). To exclude the possibility that a loss of CD2 binding capacity was simply a consequence of global disruption of the three-dimensional structure of the CD58 adhesion domain, each mutant was analyzed with a panel of eight anti-CD58 mAbs directed at four distinct native epitopes of the CD58 adhesion domain (termed epitopes 1-4) as defined by prior cross blocking studies (49). The structural integrity of the CD58 adhesion domain point mutants was assessed by reactivity with three mAbs specific for epi- CD2-CD58MolecularInteraction Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 Figure 3. Bindingof Jurkat cells to wt or mutant CD58-expressingCHO cells. CHO cells expressingwtCD58 (A) or K34A mutant CD58 (B) moleculeswere analyzedfor their abilityto interactwith Jurkat cellsand viewedunder an invertedphase-contrastmicroscope(Leitz)at a magnification of 1080. (Insets) Fluorescencehistogramsof 5,000 cells of the indicatedCHO transfectantswith anti-CD58 (RB202) (dark curve) followedby FITCconjugated goat anti-rabbit Ig or the second antibodyalone (light curve). (X-axis) Mean channelfluorescenceover a 4-log scale; (y-axis) cell number. Table 1. Summary of Effects of CD58 Mutations on CD2-CD58 Interaction CD58 Mutants Localization Levelof CD58 of Residue Expression % Binding E25A C 1.5 K30A c ~.4 i)~5: :K.3~ D~A :;;::::::::;:F: R44A CC' L L ..... :: it: 2.7 : F ::LL ::: : FF L Oil. . . . . . E72A EF ETeA F ..... anti-CD58 mAb Binding aTS2]9 IA3 IH10 IA2. IC4 IC5 IF4 PAK ~ 68 _ 5 [] I--1 [~ I--1 [] [] i'--1 I"-I 64+4 rq I--I IC3 IE3 [-1 1-3 rl I-q 26 q- 6 I-7 I--I D ~ LL:::::: :~ 79• :: : F::::::::: : FF:FFF::: : : : : : : : = : : : :::: :~:: : : 7 1"-7 I-7 ml E] I-1 I"-I [] i--I 1.3 86 +_ 7 1-7 I~ IS] I--I F-) [] [--I I-i 1.4 lo7_+4 ~ r-q El =-1 ~ [] ~ 1865 Arulanandamet al. mutants on CD2 binding to be a direct consequence of the mutated residue on CD58. CD2 Binding Site on CD58. The cellular binding analysis of 18 individual CD58 adhesion domain point mutants expressed in C H O cells identified eight charged residues on the predicted AGFCC'C" surface of the NH2-terminal CD58 adhesion domain whose mutation to alanine disrupts the binding of CD2 + Jurkat cells. The K34A C H O transfectant binds to Jurkat cells at only 1% the level of the wtCD58 C H O transfectant. In addition, the D33A, E37A, and K87A CD58 mutants show <30% binding to the CD2 + T cells. Four other CD58 point mutants (E25A, K29A, K30A, and K32A) exhibit less dramatic but nonetheless clear-cut effects on CD2 binding (35-68% wt levels). These eight residues on the adhesion domain of CD58 therefore represent potential contact sites for CD2 binding. As shown in Fig. 4, A and B, it is predicted that their side chains contribute to a charged patch covering a surface area of ~600 A 2 on the AGFCC'C" surface of CD58. It is noteworthy that this surface area is somewhat less than the 770~, z on the AGFCC'C" face of the CD2 adhesion domain known to be involved in the binding of human CD58 (43). The area is also less than the 650-690A 2 surface contact mediating the homophilic interaction involving the AGFCC'C" surfaces of two rat CD2 molecules in the rat CD2 crystal structure (34). Given that we only probed surface-exposed charged residues on CD58, it is likely that some hydrophobic and/or polar residues may participate in the CD2 binding surface of CD58, thereby contributing to a larger area of contact. For example L27, which Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 tope 1 (TS2/9, 1A3, and 1H10); one mAb specific for epitope 2 (1A2); three mAbs specific for epitope 3 (1C4, IC5, and 1F4); and one mAb specific for epitope 4 (PAK). Our results show that epitope 1 and 2 mAbs map to C', C", and G/3 strands and CC', C'C" and C"D loop regions (Fig. 2, and Table 1). Consistent with the functional evidence that epitope 1 mAbs block E rosette formation, three of the six residues (K87, E37, and K34) to which mAbs TS2/9, 1A3, and 1H10 map are involved in CD2 binding (Table 1). By way of contrast, epitopes 3 and 4 mAbs failed to block the CD58-CD2 interaction and mapped to none of the residues tested on the AGFCC'C" surface. It is thus likely that epitope 3- and 4-specific mAbs map to the BED face of the CD58 adhesion domain or alternatively, to another surface distinct from the CD2 binding region. The epitope 2 mAb cross blocks the binding of both epitopes I and 3 (49) and therefore may map to a region that straddles the major j8 strand surfaces. Our data are consistent with the mapping of 1A2 to the C"D loop (K50). Presumably, the ability of 1A2 to block CD2 binding may be mediated by steric inhibition as the excluded volume of the antibody footprint is large relative to the CD58 adhesion domain. The mutagenesis data herein provide an additional level of refinement to mAb mapping studies as they indicate that even within the epitope I group of mAbs, none react with CD58 in precisely the same way (Table 1). The differential effect of CD58 point mutants on mAb binding is evidence that mutations do not cause an overall disruption in the three-dimensional structure of the adhesion domain of CD58. Therefore, we interpret the effects of the CD58 Levels of CD58 expression for the point mutants were compared with wtCD58 and are expressedas a ratio of specificlinearmeanchannelfluorescenceintensityof Rb202 heteroantiserum binding to the mutant versus the wt. Percentbinding was calculatedby taking a ratio of totalJurkat cellbinding (averagecpm) to eachmutant relative to the wtCD58 expressingCHO cells. The standarddeviationsare shown as a percentage of the total binding. The effectof anti-CD58mAbson point mutants were comparedwith wtCD58 by taking a ratio of the relativemean fluorescence (FL) intensities of mAb binding and Rb202binding. The reactivities of anti-CD58 mAb bindingwas calculated as follows: [(FL of mAh binding to mutant - FL of control mAb bindingto mutant)/(FLof Rb202 binding to mutant - controlserabinding to mutant)]/[(FLof mAb binding to wt - FL of control mAb binding to wt)/(FL of Rb202 binding to wt - FL of control serabinding to wt)]. Reactivities <0.4 are denoted as significant effects (m), and >0.4 as no significanteffects(E]). Results are representative of a minimum of two experiments for Jurkat cell binding and for mutants that affectmAb binding. 1866 CD2-CD58 Molecular Interaction Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 1867 Arulanandam et al. Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 is predicted to be surface exposed and lies between E25, K29, K34, and E37, may be such a hydrophobic residue (Fig. 4 B). Our results are generally consistent with those of Miller et al. (17) who performed deletional mutagenesis to identify CD58 sequences involved in binding human CD2. Deletion of sequences between residues 11 and 70 on CD58 led to both a loss of CD2 and anti-CD58 mAb binding, whereas deletion of residues 131-180 in the second domain did not affect either CD2 or anti-CD58 mAb binding. However, in contrast to our finding that K87A almost completely abrogated CD2 binding, they reported that a deletion of CD58 residues 71-130 does not affect CD2 binding. Since their result is a negative one, the finding is difficult to interpret especially since another residue in the mutant structure might function in lieu of K87. Of the four CD58 residues (D33, K34, E37, and K87) that apparently play the greatest role in CD2 binding (<30% wt binding), K34 is conserved among human and sheep CD58, and human CD48 and human CD2 (Fig. 1, and Table 1). Similarly, E37 is conserved in human CD48 and the comparable residue in sheep CD58 and human CD2 is the related 1868 CD2-CD58MolecularInteraction Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 Figure 4. Delineationof the CD2 binding site on CD58 and possible CD2/CD58 docking orientations. The CD2 binding site on CD58 is a highly charged surfacearea on the predictedAGFCC'C" surfaceof the adhesiondomainof CD58. Shownis a stereo or-carbontrace model (A) and a correspondingspace-fillingmodel(B). Possibledockingorientations for the CD2 (white) and CD58 (blue) interactionsbased on the rat CD2 dimer(C) or CD8 ot homodimer(/9). A space-fillingmodelfor the docking orientationis shown(E). Residueson CD58 that showed<30% (red)(D33, K34, F.37, and K87) or 30-75% (yellow) (E25, K29, K30, and K32) of wt binding to CD2 when substitutedwith alanineare indicated.PreviouslydefinedCD2 residuesaffectingCD58 binding(43) are alsodepicted. The COOH-terminal segment of the G strand of the CD58 and CD2 adhesion domain is indicatedby an open box. amino acid glutamine. K87 is also conserved between human CD58 and CD48. In view of the conservation of K34, E37, and K87 in human CD58 and CD48 and their observed functional effects on the CD2-CD58 interaction, it is likely that the corresponding residue in CD48 is involved in the CD2 interaction. If this is so, CD48 and CD58 may bind to CD2 in a similar way. In contrast, the other five charged CD58 residues (E25, K29, K30, K32, and D33) which contribute to CD2 binding, are conserved between human and sheep variants of CD58 but not human CD48. These residues may, at least in part, be responsible for the higher affinity of the CD58 interaction with human CD2 (Kd ~10-6 M) relative to the human CD48 interaction with human CD2 (Kd r~10 -4 M). Polar and/or hydrophobic residues yet to be identified may also contribute to the higher-affinity interaction between CD2 and CD58. Collectively, these data also suggest that human and sheep CD58 interact with CD2 in a similar manner, consistent with our prior analysis of mutant CD2 molecules (43). In that study, it was observed that CD2 mutations which disrupted the sheep CD58 interaction also disrupted the human CD58 interaction. Possible Docking Orientation of CD58 and CD2. The mutational data for CD58 presented herein and for CD2 presented elsewhere (43) define the area of interaction between these proteins. To develop a model of the CD2-CD58 complex, we considered as paradigms several examples of crystaUographically defined docking orientations between Ig-related molecules. These include the known interaction between Ig VH and VL domains, between human CD8 c~ subunits in the CD8 or-or homodimer and between two CD2 molecules in the rat CD2 crystal structure (34, 46, 47). Since the packing between CD8 ot subunits is close to that between V. and V~, the latter were not studied separately. However, the orientation of molecules in the rat CD2 dimer structure is distinct from that of CD8 or Ig dimers (34). In the CD8 ~-t~ homodimer structure, the molecular association between subunits involves main chain hydrogen bonding between the C' strand of one molecule and the G strand of the other. In the rat CD2 dimer, the CC' and FG loops are shortened compared to CD8, and the relative orientations of their domains differ by ~60 ~ from that of CDS. This reduces the extent of the hydrogen bonding interaction to the beginning of the C' and G strands where they join the CC' and FG loops, respectively. Based on these examples, two possible docking orientations for the CD58-CD2 interaction were created. The first docking orientation (Fig. 4 C) was constructed using a model of the rat CD2 dimer: one rat CD2 adhesion domain was replaced by the human CD2 NMR structure by least squares superimposition of the C~ atoms of residues in the binding sheet and the other domain substituted by the human CD58 model. As shown in Fig. 4 C, this orientation brings the FG and CC' loop of CD2 and CD58 adjacent to each other. In this docking orientation, the COOH-terminal end of the G strands of CD2 and CD58 point away from each other in essentially opposite directions, presumably towards two separate cell membranes. Fig. 4 D represents a second pos- CD2 binding is yet to be determined. The ability of sulfated dextrans to inhibit the CD2-CD58 interaction is probably a consequence of direct binding of these polymers to the positively charged CD2 and/or CD58 surface (50). The mapping of the CD2 binding site to the predicted AGFCC'C" surface of the CD58 adhesion domain shows that the major/3 sheet surfaces mediating interaction between CD2 and CD58 are the same as those mediating dimerization of other Ig-variable domains including Ig, CD8c~, and rat CD2 (34, 46, 47). In the CD8ot homodimer and V,-VL interactions, the G and C' strands contain conserved/3 bulges located at the dimer interface which are believed to facilitate dimerization (47, 51, 52). The CD58 residue E37 on the predicted C' strand and residues K32, D33, and K34 adjacent to the start of this strand and K87 within the predicted G strand are all important for CD2 binding. These residues may actually be components of equivalent 3 bulges. Support for this notion is the finding that the corresponding residue in the CD2 C' strand (Q46) and residues adjacent to this (K41 and K43) are critical for CD58 binding (43, 53). They are located on the corresponding/3 bulge of CD2 (32, 33). Mutations of G90, K91, N92, and V93 near the start of the CD2 G strand disrupt CD58 binding but there is no evidence for a 3 bulge in the G strand (33). The prevalence of charged residues in the predicted CD2-CD58 interface noted above appears to differ from the predominant hydrophobic interaction among residues found at the site of VH-VL domain packing (51). However, if there is a 3 bulge located in the G strand of CD58 involving residues K87, F88, and F89, then this bulge would expose hydrophobic side chains that would involve hydrophobic contacts between CD2 and CD58. The present analysis of the molecular interface in the CD2-CD58 complex offers a first consideration of the nature of a receptor-ligand pair involving two Ig supergene family members. Such interactions are distinct from those between V domains in idiotype-anti-idiotype complexes where major surface contacts are mediated via the CDR1 (BC), CDR2 (C'C"), and CDR3 (FG) loops (54, 55). They are also different from the interaction between Ig constant region domains which involve the BED/3 sheet surface (57). Given that either of the two CD2-CD58 docking orientations might facilitate conjugate formation between T lymphocytes and APCs, the precise orientation of the interaction awaits elucidation by structural analysis of a CD2-CD58 complex. Alternatively, it may be possible to infer the correct docking mode by further mutagenesis involving reciprocal charge reversals in CD2 and CD58. We thank Dr. Tetsu Kakutani for supplying the sheep CD58 sequence prior to submission to the DDBJ DNA data base; Dr. Laurelee Osborn for providing the Rb202 anti-CD58 hetero anti-sera; Drs. Stefan Meuer and Ulrich Moebius for providing anti-CD58 mAbs; Dr. Reinhard Wallich for sharing unpublished information on genomic cloning of CD58; and Dr. Per Kraulis for the "MOLSCRIPT" program. We also acknowledge Ms. Yasmin Hussain for oligonucleotide synthesis and Mr. Peter Lopez and members of the core flow cytometry facility for flow cytometric analysis. 1869 Arulanandamet al. Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 sible docking orientation based on the coordinates of the CD8 c~ homodimer which bring the FG loops of CD2 and CD58 near one another as shown. If this orientation is correct, then the COOH-terminal segments of the G strand in CD2 and CD58 are only separated by an angle of ~60 ~ Given that the crystal structure of the two domain rat CD2 molecule identifies an angle of ~160 ~ between the adhesion domain of CD2 and the second domain plus COOH-terminal stalk region (34), it is plausible that the adhesion domains of CD2 and CD58 could approach one another from two separate cell membranes in such a manner. Both docking models take into account all the side chains implicated in binding with one exception. In the CD2 dimer-based model (Fig. 4 C), K87 on CD58, which is involved in CD2 binding, is an outlier and makes no contacts with CD2. In contrast, in the CDSot-based model (Fig. 4 D), it is on the edge of the binding interface and can be involved in direct interaction. This difference tends to support the CD8ot-based model as more representative of the real CD2-CD58 docking orientation. On the other hand, our efforts at de novo docking of CD2 and CD58 suggest an orientation close to the CD2 dimer-based model that readily provides K87 with contacts to CD2 (data not shown). Hence, we cannot make a definitive statement about the docking orientations of CD2 and CD58 at present. A space-filling model of the CD2-CD58 interaction in the CD8-based orientation (Fig. 4 E) shows that the nature of the contact surface size and shape is also reasonable. Implications for the CD2-CD58 Interaction. Our prior mutagenesis analysis shows that the CD58 binding site on CD2 covers --770A 2 on the AGFCC'C" face of the CD2 ~ barrel (43). This site contains 3 strand residues in the C O O H terminal half of the F strand (including K82 and Y86), the top of the C strand (D32 and K34), and the C' strand (Q46), which are all solvent exposed. In addition, several exposed residues on the FG loop (G90, K91, N92, and V93), the CC' loop (K41 and K43) and the C'C" loop (R48 and K51) form this site. The current mutagenesis on CD58 has uncovered the other surface of the interaction involving five positively charged residues (K29, K30, K32, K34, and K87) and three negatively charged residues (E25, D33, and E37). In light of the seven positively charged and one negatively charged CD2 residues involved in this site, it appears there is an excess of positive charge. This excess positive charge may be satisfied by interaction with water. Alternatively, the positive charge may also be satisfied by negatively charged side chains that appear not to be crucial for binding (E39, E76, and E78) or are near the site (E42 and E74) but whose contribution to This work was supported by National Institutes of Health grants AI-21226 to E. L. Reinherz and GM38608 to G. Wagner. Address correspondence to Dr. A. R. N. Arulandandam, Laboratory of Immunobiology, Dana-Farber Cancer Institute, 44 Binney Street, Boston, MA 02135. Received for publication 25 May 1994 and in revised form 15July 1994. 1870 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. CD2-CD58 Molecular Interaction Malissen, J. Kochan, H.-C. Chang, H.-R. Rodewald, and E.L. Reinherz. 1992. CO3 ~"dependence of the CO2 pathway of activation in T lymphocytes and natural killer cells. Proa Natl. Acad. Sci. USA. 89:1492. Sayre, P.H., R.E. Hussey, H.-C. Chang, T.L. Ciardelli, and E.L. Reinherz. 1989. Structural and binding analysis of a two domain extracellular CO2 molecule. J. Exl~ Med. 169:995. Recny, M.A., E.A. Neidhardt, P.H. Sayre, T.L. Ciardelli, and E.L. Reinherz. 1990. Structural and functional characterization of the CO2 immunoadhesion domain. J. Biol. Chem. 265:8542. Miller, G.T., P.S. Hochman, W. Meier, R. Tizard, S.A. Bixler, M.D. Rosa, and B.P. Wallner. 1993. Specific interaction of lymphocyte function-associated antigen 3 with CO2 can inhibit T cell responses. J. Exlx Ivied. 178:211. Moingeon, P., J.L. Lucich, C.C. Stebbins, M.A. Recny, B.P. Wallner, S. Koyasu, and E.L. Reinherz. 1991. Complementary roles for CO2 and LFA-1 adhesion pathways during T cell activation. Eur.J. ImmunoI. 21:605. Tiefenthaler, G., M.L. Dustin, T.A. Springer, and T. Hunig. 1987. Serologic crossreactivity ofT11 target structure and lymphocyte function-associated antigen 3.J. Immunol. 139:2696. Lay,W.H., N.F. Mendes, C. Bianco, and V. Nussenzweig. 1971. Binding of sheep red blood cells to a large population of human lymphocytes. Nature (Lond.). 230:531. Brain, P., J. Gordon, and W.A. Willetts. 1970. Rosette formation by peripheral lymphocytes. Clin. EXlXImmunol. 6:681. Coombs, R.R.A., R.W. Burner, A.B. Wilson, G. Holm, and B. Lindgren. 1970. Rosette-formation between human lymphocytes and sheep red cells not involving immunoglobulin receptors. Int. Arch. Allergy Appl. Immunol. 39:658. Jondal, M., G. Holm, and H. Wigzell. 1972. Surface markers on human T and B lymphocytes: a large population of lymphocytes forming nonimmune rosettes with sheep red blood cells, j. Extz Med. 136:207. Arulanandam, A.R.N., P. Moingeon, M.F. Concino, M.A. Recny, K. Kato, H. Yagita, S. Koyasu, and E.L. Reinherz. 1993. A soluble multimeric recombinant CO2 protein identifies CD48 as a low affinity ligand for human CO2: divergence of CO2 ligands during the evolution of humans and mice.J. Extx Med. 177:1439. Kato, K., M. Koyanagi, H. Okada, T. Takanashi, Y.-W. Wong, A.F. Williams, K. Okumura, and H. Yagita. 1992. CD48 is a counter-receptor for murine CO2 and is involved in T cell activation. J. Exft Med. 176:1241. Sayre, P.H., H.-C. Chang, R.E. Hussey, N.R. Brown, N.E. Richardson, G. Spagnoli, L.K. Clayton, and E.L. Reinherz. 1987. Molecular cloning and expression ofT11 cDNAs reveal a receptor-like structure on human T lymphocytes. Proa Natl. Acad. Sci. USA. 84:2941. Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 ReFerences 1. Moingeon, P., H.-C. Chang, P.H. Sayre, L.K. Clayton, A. Alcover, P. Gardner, and E.L. Reinherz. 1989. The structural biology of CO2. Immunol. Rev. 111:111. 2. Springer, T.A. 1990. Adhesion receptors of the immune system. Nature (Lond.). 346:424. 3. Bierer, B.E., B.P. Sleckman, S.E. Ratnofsky, and S.J. Burakoff. 1989. Biologic roles of CO2, CO4 and CO8 in T cell activation. Annu. Rev. Immunol. 7:579. 4. Moingeon, P., H.-C. Chang, B.P. Wallner, C. Stebbins, A.Z. Frey, and E.L. Reinherz. 1989. CD2-mediated adhesion facilitates T lymphocyte antigen recognition function. Nature(Lond.). 339:312. 5. Koyasu, S., T. Lawton, D. Novick, M.A. Recny, R.F. Siliciano, B.P. Wallner, and E.L. Reinherz. 1990. Role of interaction of CO2 molecules with lymphocyte function associated antigen 3 in T cell recognition of nominal antigen. Proa Natl. Acad. Sci. USA. 87:2603. 6. Meuer, S.C., R.E. Hussey, M. Fabbi, D. Fox, O. Acuto, K.A. Fitzgerald, J.C. Hodgdon, J.P. Protentis, S.F. Schlossman, and E.L. Reinherz. 1984. An alternative pathway of T cell activation: a functional role for the 50 KD Tll sheep erythrocyte receptor protein. Cell. 36:897. 7. Yang, S.-Y., S. Chouaib, and B. Dupont. 1986. A common pathway for T lymphocyte activation involving both the CD3-Ti complex and CO2 sheep erythrocyte receptor determinants. J. Immunol. 137:1097. 8. HOnig, T., G. Tiefenthaler, K.-H. Meyer zum BOschenfelde, and S.C. Meuer. 1987. Alternative pathway activation of T cells by binding of CO2 to its cell surface ligand. Nature (Lond.). 326:298. 9. Bierer, B.E., A. Peterson, J.C. Gorga, S.H. Herrmann, and S.J. Burakoff. 1988. Synergistic T cell activation via the physiological ligands for CO2 and the T cell receptor.J. Exp. Med. 168:1145. 10. Chang, H.-C., P. Moingeon, P. Lopez, H. Krasnow, C. Stebbins, and E.L. Reinherz. 1989. Dissection of the human CO2 intracellular domain. Identification of a segment required for signal transduction and interleukin 2 production. J. Exp. Med. 169:2073. 11. He, Q., A.D. Beyers, A.N. Barclay, and A.F. Williams. 1988. A role in transmembrane signalling for the cytoplasmic domain of the CO2 T lymphocyte surface antigen. Cell. 54:979. 12. Siliciano, R.F., J.C. Pratt, R.E. Schmidt, J. Ritz, and E.L. Reinherz. 1985. Activation of cytolytic T lymphocytes and natural killer cell function through the Tll sheep erythrocyte binding protein. Nature (Lond.). 317:428. 13. Arulanandam, A.R.N., S. Koyasu, and E.L. Reinherz. 1991. T cell receptor-independent CO2 signal transduction in FcR § cells. J. Exp. Med. 173:859. 14. Moingeon, P., J.L. Lucich, D.J. McConkey, F. Letourneur, B. 1871 Arulanandam et al. and D.A. Thorley-Lawson. 1991. Expression of the Blast-1 activation/adhesion molecule and its identification as CD48. J. Immunol. 146:2192. 42. Reinherz, E.L., and S.F. Schlossman. 1980. The differentiation and function of human T lymphocytes. Cell. 19:821. 43. Arulanandam, A.R.N., J.M. Withka, D.F. Wyss, G. Wagner, A. Kister, P. Pallai, M.A. Recny, and E.L. Reinherz. 1993. The CD58 (LFA-3) binding site is a localized and highly charged surface area on the AGFCC'C" face of the human CD2 adhesion domain. Proc. Natl. Acad. Sci. USA. 90:11613. 44. Somoza, C., PC. Driscoll, J.G. Cyster, and A.F. Williams. 1993. Mutational analysis of the CD2/CD58 interaction: the binding site for CD58 lies on one face of the first domain of human CD2. J. Exp. Med. 178:549. 45. Dengler, T.J., J.C. Hoffmann, P. Knolle, M.A.-Wolf, M. Koux, K. Wallich, and S.C. Meuer. 1992. Structural and functional epitopes of the human adhesion receptor CD58 (LFA-3). Eur. J. Immunol. 22:2809. 46. Epp, O., E.E. Lattman, M. Schiffer, and K. Huber. 1975. The molecular structure of a dimer composed of the two variable portions of the Bence-Jones protein/REI. Biochemistry. 14:4943. 47. Leahy, D.J., R. Axel, and W. Hendrickson. 1992. Crystal structure of a soluble form of the human T cell coreceptor CD8 at 2.6 fi~ resolution. Cell. 68:1145. 48. Wang, J., Y. Yan, T.P.J. Garrett, J. Liu, D.W. Rodgers, R.L. Garlick, G.E. Tart, Y. Hussain, E.L. Reinherz, and S.C. Harrison. 1990. Atomic structure of a fragment of human CD4 containing two immunoglobulin-like domains. Nature (Lond.). 348:411. 49. Gray, F., J.G. Cyster, A. Willis, A.N. Barclay, and A.F. Williams. 1993. Structural analysis of the CD2 T lymphocyte antigen by site-directed mutagenesis to introduce a disulphide bond into domain 1. Protein Eng. 6:965. 50. Parish, C.K., V. McPhun, and H.S. Warren. 1988. Is a natural ligand of the T lymphocyte CD2 molecule a sulfated carbohydrate? J. Immunol. 141:3948. 51. Chothia, C., J. Novotny, R. Bruccoleri, and M. Karplus. 1985. Domain association in immunoglobulin molecules: the packing of variable domains. J. Mol. Biol. 186:651. 52. Colman, P.M. 1988. Structure of antibody-antigen complexes: implications for immune recognition. Adv. Immunol. 43:99. 53. Peterson, A., and B. Seed. 1987. Monoclonal antibody and ligand binding sites of the T cell erythrocyte receptor (CD2). Nature (Lond.). 329:842. 54. Bentley, G.A., G. Boulot, M.M. Riottot, and K.J. Poljak. 1990. Three-dimensional structure of an idiotope-anti-idiotope complex. Nature (Lond.). 348:254. 55. Ban, N., C. Escobar, R. Garcia, K. Hasel, J. Day, A. Greenwood, and A. McPherson. 1994. Crystal structure of an idiotype-anti-idiotype Fab complex. Proa NatL Acad. Sci. USA. 91:1604. 56. Kraulis, P.J. 1991. MOLSCKIPT: a program to produce both detailed and schematic plots of protein structures. J. Applied Crystallography. 24:946. 57. Edmunson, A.B., K.K. Ely, E.E. Abola, M. Schiffer, and N. Panagiotopoulous. 1975. Rotational aUomerism and divergent evolution of domains in Ig light chains. Biochemistry. 14:3953. Downloaded from http://rupress.org/jem/article-pdf/180/5/1861/1676597/1861.pdf by guest on 26 July 2023 27. Seed, B., and A. Aruffo. 1987. Molecular cloning of the CD2 antigen, the T cell erythrocyte receptor, by a rapid immunoselection procedure. Proa Natl. Acad. Sci. USA. 84:3365. 28. Seed, B. 1987. An LFA-3 cDNA encodes a phospholipid-linked membrane protein homologous to its receptor CD2. Nature (Lond.). 329:840. 29. Wallner, B.P., A.Z. Frey, K. Tizard, R.J. Mattaliano, C. Hession, M.E. Sanders, M.L. Dustin, and T.A. Springer. 1987. Primary structure of lymphocyte function-associated antigen (LFA-3): the ligand of the T lymphocyte CD2 glycoprotein. J. Extx Med. 166:923. 30. Williams, A.F., and A.N. Barclay. 1987. The immunoglobulin superfamily - domains for cell surface recognition. Annu. Rev. Immunol. 6:381. 31. Driscoll, P.C., J.G. Cyster, I.D. Campbell, and A.F. Williams. 1991. Structure of domain 1 of rat T lymphocyte CD2 antigen. Nature (Lond.). 353:762. 32. Withka, J.M., D.F. Wyss, G. Wagner, A.R.N. Arulanandam, E.L. Reinherz, and M.A. Recny. 1993. Structure of the glycosylated adhesion domain of human T lymphocyte glycoprotein CD2. Structure. 1:69. 33. Wyss, D.F.,J.M. Withka, M.H. Knoppers, K.A. Sterne, M.A. Recny, and G. Wagner. 1993. 1H Resonance assignments and secondary structure of the 13.6 KDa glycosylated adhesion domain of human CD2. Biochemistry. 32:10995. 34. Jones, E.Y., S.J. Davis, A.F. Williams, K. Harlos, and D.I. Stuart. 1992. Crystal structure at 2.8fix resolution of a soluble form of the cell adhesion molecule CD2. Nature (Lond.). 360:232. 35. Staunton, D.E., and D.A. Thorley-Lawson. 1987. Molecular cloning of the lymphocyte activation marker Blast-1. EMBO (Fur. Mol. Biol. Organ.) J. 6:3695. 36. Staunton, D.E., K.C. Fisher, M.M. LeBeau, J.B. Lawrence, D.E. Barton, U. Francke, M. Dustin, and D.A. ThorleyLawson. 1989. Blast-1 possesses a glycosyl-phosphatidylinositol (GPI) membrane anchor, is related to LFA-3 and OX-45, and maps to chromosome lq21-23. J. Exp. Med. 169:1087. 37. Sewell, W.A., K.W. Palmer, N.K. Spurr, D. Sheer, M.H. Brown, Y. Bell, and M.J. Crumpton. 1988. The human LFA-3 gene is located at the same chromosome band as the gene for its receptor CD2. Immunogenetics. 28:278. 38. Wong, Y:W., A.F. Williams, S.F. Kingsmore, and M.F. Seldin. 1990. Structure, expression, and genetic linkage of the mouse BCM1 (OX45 or Blast-l) antigen: evidence for genetic duplication giving rise to the BCM1 region on mouse chromosome 1 and the CD2/LFA3 region on mouse chromosome 3.J. Exp. Med. 171:2115. 39. Diamond, D.J., L.K. Clayton, P.H. Sayre, and E.L. Reinherz. 1988. Exon-intron organization and sequence comparison of human and murine Tll (CD2) genes. Pro~ Natl. Acad. &i. USA. 85:1615. 40. Krensky, A.M., F. Sanchez-Madrid, E. Robbins, J.A. Nagy, T.A. Springer, and S.J. Burakoff. 1983. The functional significance, distribution and structure of LFA-1, LFA-2 and LFA-3: cell surface antigens associated with CTL-target interactions. J. Immunol. 131:611. 41. Yokoyama, S., D. Staunton, R. Fisher, M. Amiot, J.J. Forin,