2B4, the Natural Killer and T Cell Immunoglobulin Superfamily Surface Protein, Is a Ligand for CD48

mAbs used were 2B4 (10); rat (r)CD4d3+4 mAb, OX68; mouse (m)CD2 mAb, RM2.1 (19); mCD48 mAb, OX78; rCD48 mAb, OX45; and rCD147 mAb, OX47. The OX mAbs are all referenced in the European Collection of Animal Cell Cultures (Porton Down, Wiltshire, UK). Purified HM48-1 (hamster anti–mouse CD48 [5]) was provided by Dr. H. Yagita (Juntendo University School of Medicine, Tokyo, Japan), and control 2C11 (hamster anti–mouse CD3) by Dr. E. Adams (Sir William Dunn School of Pathology). Human (h)CD48 mAbs were MEM 102 (20) and 6.28 (21), provided by Drs. V. Horejsi (Institute of Molecular Genetics, Prague, Czech Republic) and D. Thorley-Lawson (Department of Pathology, Tufts University, Boston, MA), respectively.

The constructs for CD4d3+4 chimeric proteins used previously (22) were modified to contain at the COOH terminus a consensus peptide sequence recognized by the Escherichia coli biotin holoenzyme synthetase, BirA (17, 18). DNA encoding the amino acid sequence NSGSLHHILDAQKMVWNHR* was amplified by PCR using 5′ (agggttgaattccggatcactgcatcatattctgg) and 3′ (ctactaggatccttaacgatgattccacacc) oligonucleotides (restriction enzyme sites are underlined) and an HLA A2 construct as template (17), provided by Dr. Paul Moss (Institute of Molecular Medicine, Oxford, UK). This fragment was inserted via a 5′ EcoRI site and a 3′ BamHI site into a modified Bluescript (BS KS⁺) vector containing a rat CD4 leader and rat CD4d3+4 (CD4Ld3+4RI [23]). An XbaI-BamHI fragment encoding XbaI-CD4L-SalI-CD4d3+4-EcoRI-biotin peptide-stop-BamHI was transferred to a modified pEF-BOS vector (24), pEF-BOS-XB. To construct fusion proteins, the XbaI-SalI fragment encoding CD4L was replaced with DNA encoding the extracellular region of m2B4 or h2B4 amplified from plasmid DNA (13; Boles, K., and P.A. Mathew, manuscript in preparation) or domains 1–3 of rCD5 (25). XbaI sites were introduced 42 bases (m2B4) and 52 bases (h2B4) upstream of the initiation Met. The join at the SalI (g tcg acc) junction with rCD4d3 was m2B4:VPSNFRST (the residues containing the SalI site are underlined) and at the homologous region for h2B4. Confirmatory sequence analysis of all constructs (model 373A DNA sequencing system; Applied Biosystems, Inc., Foster City, CA) revealed for m2B4-CD4d3+4 a 3-base deletion, the correct sequence reading at base 619: gctttgtac encoding ALY in the C strand of domain 2. To construct soluble His-tagged hCD48, DNA (sequence data available from EMBL/GenBank/ DDBJ under accession no. M37766) encoding the extracellular region with an XbaI site 13 bases upstream of the initiation Met and a His-tag:TLARSTHHHHHH* contained in the 3′ oligonucleotide (ccactaggatcctaatgatggtggtgatgatgggtcgaccgggccagggtacagggtggactgag) was amplified by PCR from mouse macrophage cDNA (provided by Dr. J. Mahoney, Sir William Dunn School of Pathology) and cloned into pEF-BOS-XB.

Recombinant proteins were expressed by transfection of 293T cells with 40 μg plasmid DNA/5 × 10⁶ cells/175-cm² flask using calcium phosphate. After removal of precipitate at 16 h, cells were incubated for 48 h in 7.5–10 ml X-Vivo-10 serum-free medium (BioWhittaker, Walkersville, MD). Levels of four Ig domain fusion proteins as assayed in a CD4d3+4 inhibition ELISA (22) were routinely in the region of 50–150 μg/175-cm² flask. To enzymatically biotinylate CD4d3+4 fusion proteins, supernatant was exchanged (∼1:20) into 10 mM Tris-HCl, pH 8, and concentrated to 0.5–0.8 ml using a 10,000 mol wt cutoff 15-ml centricon (Amicon, Inc., Beverly, MA) and incubated with 1 μl BIR enzyme (Avidity, Denver, CO). Biotinylation was complete after 2–4 h at 30°C or overnight incubation at room temperature as assessed by binding of streptavidin to chimeras immobilized via a CD4d3+4 mAb in a BIAcore™ (reference 26, and see below). Excess biotin was removed by dialysis (twice with 800 ml PBS for 20 h), and biotinylated material was aliquoted and stored at −20°C.

Purified monomeric mCD48 containing a COOH-terminal His-tag was prepared as described (27). Soluble hCD48 was expressed as for CD4d3+4 fusion proteins in DMEM containing 10% FCS. Soluble hCD48 was purified using an FPLC (Pharmacia Biotech Ltd., St. Albans, UK) first on a 1-ml HiTrap^® column (Pharmacia Biotech Ltd.). Monomeric soluble hCD48 was then eluted in 0.1-ml fractions from Superdex 75 (Pharmacia Biotech Ltd.) and used within 4 h. Concentration was measured at 280 nm using an extinction coefficient of 1.6 cm²/mg (28), and purity was checked by SDS-PAGE. Purified hCD48 was shown to be antigenically active by binding stoichiometric amounts of the 6.28 mAb when immobilized via the His-tag on an NTA chip (Biacore AB, Uppsala, Sweden). In experiments with h2B4-CD4d3+4 and control CD4d3+4 immobilized on streptavidin-coated beads (Dynal A.S., Oslo, Norway), >70% hCD48 could be depleted as assessed by SDS-PAGE and densitometry.

Experimental procedures were essentially as described (16). Biotinylated proteins (1–2 μg/sample) were mixed with 10–20 μl avidin-coated fluorescent beads/sample (VFP-0552-5; Spherotech, Inc., Libertyville, IL). Small volumes were shaken in round-bottomed microtiter plates. Where volume adjustment was necessary, beads were centrifuged and resuspended in PBS containing 0.2% BSA (PBS/BSA). For coating beads with mAbs in blocking studies, beads were washed twice in PBS/BSA and resuspended in PBS/BSA containing 2B4 or OX68 mAb (1 μg/20 μl beads).

Experiments were carried out on a BIAcore™ 2000 (Biacore AB) using HBS buffer (25 mM Hepes, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, and 0.005% surfactant P20) supplied by the manufacturer. Streptavidin 0.2 mg/ml (Pierce Chemical Co., Rockford, IL) was coupled in 10 mM sodium acetate, pH 5, to a research grade CM5 chip (Biacore AB) using an amine coupling kit (Biacore AB) using an activation time of 5 min, resulting in immobilization of ∼2,000–3,000 response units (RU). Soluble mCD2 comprising the extracellular region of mCD2 (27) at 90 μg/ml was coupled similarly with an activation time of 1.8 min. The surface of the chip was washed with 0.1 M glycine-HCl, pH 2.5, after coupling. Biotinylated 2B4-CD4d3+4 and control rCD5-CD4d3+4 or CD4d3+4 were then injected over immobilized streptavidin. Equilibrium affinity measurements were performed as described (26). Increasing and decreasing concentrations of mCD48 (5-μl injections at 20 μl/min) were passed over m2B4-CD4d3+4, rCD5-CD4d3+4, and mCD2. Kinetic measurements were made at 37°C (28). mCD48 (23.4 μM) was injected for 3 s at 100 μl/min over high and low levels of m2B4-CD4d3+4, rCD5-CD4d3+4, and mCD2. Similarly, to measure affinity, hCD48 was passed over h2B4-CD4d3+4 and CD4d3+4, and for kinetics, hCD48 (14.4 μM) was passed over high and low levels of h2B4-CD4d3+4 and CD4d3+4. Dissociation rates were measured for mCD48 as described (28) using Origin software, version 5.0 (Microcal Software, Inc., Northampton, MA). The effect of 2B4 mAb on mCD48 (23.4 μM) binding was tested by first passing 2B4 mAb at 10 μg/ml at 5 μl/min over m2B4-CD4d3+4 until saturation was reached.

A binding reagent for a m2B4 ligand was constructed by expressing the two IgSF domains of 2B4 as a recombinant chimeric protein with domains 3 and 4 of rCD4 (m2B4-CD4d3+4; Fig. 1,B). The recombinant protein differed from previously described CD4d3+4 chimeras (22) in that it contained a peptide sequence at its COOH terminus that can be specifically biotinylated by the E. coli enzyme, BirA (18). The expressed protein was biotinylated by the BirA enzyme in vitro and tested for reactivity with streptavidin and 2B4 mAb by surface plasmon resonance (not shown). The in vitro biotinylation of the recombinant protein simplified the previously described method (16) for detecting low-affinity interactions at the cell surface by enabling the biotinylated m2B4-CD4d3+4 to be directly immobilized on avidin-coated beads (Fig. 1 C) without the use of a CD4d3+4 mAb. The assay was shown to be at least as sensitive as coupling CD4d3+4 fusion proteins to beads using a CD4d3+4 mAb (data not shown).

m2B4-CD4d3+4 beads bound to LPS and Con A–activated mouse spleen cells (Fig. 2, A and B). The broad distribution of the ligand and our own expectation that it might be another member of the CD2 subset suggested CD48 as a likely candidate. In support of this, the binding of m2B4-CD4d3+4 beads was blocked by the CD48 mAb, OX78, but not by a control CD2 mAb (Fig. 2, A and B). A second CD48 mAb, HM48-1, which blocks the CD2–CD48 interaction (5) and is known to have functional effects in vivo (29, 30), also inhibited the binding of m2B4-CD4d3+4 beads (Fig. 2 C).

Preincubation of m2B4-CD4d3+4 beads with 2B4 mAb before presentation to cells partially inhibited binding, whereas a control CD4d3+4 mAb had no effect (Fig. 2 D). The partial inhibition suggests that sites recognized by 2B4 mAb and CD48 may be close but do not overlap.

The m2B4-CD4d3+4 beads also bound to rat thymocytes (not shown) and a rat mast cell line (Fig. 2,E). This binding was completely inhibited by a rat CD48 mAb (Fig. 2 E), indicating that the 2B4–CD48 interaction is conserved between these species, as is the CD2–CD48 interaction (27).

The finding that rodent 2B4 and CD48 bind to each other suggests that 2B4 may be a ligand for CD48 in humans. The recent isolation of h2B4 cDNA (Boles, K., and P.A. Mathew, manuscript in preparation) made it possible to test this hypothesis. An h2B4 rCD4d3+4 fusion protein with a COOH-terminal biotinylation tag (h2B4-CD4d3+4) was expressed at levels similar to those obtained with the m2B4-CD4d3+4 construct, consistent with it being correctly folded. Biotinylated h2B4-CD4d3+4 immobilized on fluorescent beads bound to PBMCs (Fig. 3,A). Two hCD48 mAbs (6.28 and MEM 102) were tested for blocking effects on h2B4-CD4d3+4 cell binding. Preincubation of cells with the mAbs revealed blocking of h2B4-CD4d3+4 binding by 6.28 mAb but not by MEM 102 mAb (Fig. 3,A). Both 6.28 and MEM 102 mAbs bound to the cells, with the IgG3 6.28 mAb giving a lower signal as detected by the FITC-labeled secondary reagent (Fig. 3 B). The specificity of blocking by a CD48 mAb against a particular epitope provides convincing evidence that antigenically active h2B4 is binding to CD48 at the cell surface.

The finding that mCD48 has two ligands, 2B4 and CD2, raises the question of which interaction dominates in a particular response. This will depend on the distribution and level of expression of the proteins and the affinities of the two interactions. The latter was investigated by analysis on a BIAcore™, which permits protein interactions to be followed in real time using a detection system based on surface plasmon resonance. The affinity was measured by determining the level of binding at equilibrium after injection of a range of concentrations of mCD48 over immobilized m2B4-CD4d3+4 and mCD2. Immobilized rCD5-CD4d3+4 was used as a negative control in the reference flow cell. A representative experiment with increasing concentrations of mCD48 injected at 37°C is shown in Fig. 4,A. The difference between the response at equilibrium in the m2B4 and control flow cells represents binding (Fig. 4,B). A Scatchard transformation of the data is shown in Fig. 4,C. mCD48 bound to m2B4 with a K_d ≈ 16 μM at 37°C (Fig. 4, B and C, and Table 1). This was significantly higher than the mCD48–mCD2 interaction (K_d ≈ 90 μM) measured simultaneously (Fig. 4,C, and Table 1). Values determined for the mCD48–mCD2 interaction agree with published data (27; Table 1). At 25°C, mCD48 bound m2B4 and mCD2 with a K_d of 7 and 49 μM, respectively (data not shown).

Kinetic studies were carried out at 37°C to determine the dissociation rate constant (k_off) of the mCD48–m2B4 interaction. mCD48 dissociated more slowly from m2B4 (k_off = 3 s⁻¹) than from mCD2 (k_off >20 s⁻¹) (Fig. 4 D). The k_off was the same when the level of immobilized m2B4-CD4d3+4 was decreased, indicating that the dissociation rate measured was not affected by rebinding or mass transport limitations. This suggests that the higher affinity of the 2B4–CD48 interaction is the result of a decreased k_off.

The partial blocking effect of 2B4 mAb on cell binding (Fig. 2,D) was reproduced in BIAcore™ analysis as shown in Fig. 4 E. Saturation of m2B4-CD4d3+4 with 2B4 mAb before injection of mCD48 reduced the level of mCD48 binding.

To measure the affinity of the interaction between h2B4 and hCD48, a soluble form of hCD48 was produced. In analyses similar to that carried out for the m2B4–mCD48 interaction, affinity measurements were made with soluble hCD48 and immobilized h2B4-CD4d3+4. Both nonlinear curve fitting (Fig. 5,B) and linear regression analysis of the Scatchard plot (Fig. 5,C) yielded a K_d ≈ 8 μM (Table 1). The apparent k_off was 5 and 7 s⁻¹ when measured with high and low levels of immobilized h2B4 (Fig. 5 D). This indicates that dissociation is limited by mass transport and/or rebinding and that the true k_off is ≥7 s⁻¹.

The interaction between 2B4 and CD48 identified in this study represents a new ligand pair that is conserved between rodents and humans within the CD2 subset of IgSF molecules. A previous study failed to demonstrate an additional ligand for CD48 in the rat (16). However, 2B4 is expressed on a limited range of cell types, and these were not examined (16). Interestingly, in preliminary experiments, the binding of biotinylated multivalent CD48-CD4d3+4 to lymphokine-activated killer (LAK) cells shown to express both 2B4 and CD2 was entirely blocked by a CD2 mAb (Brown, M.H., unpublished observations). It is possible that with the particular conditions used to generate the LAK cells, too few expressed 2B4 at a sufficiently high surface density and/or 2B4 has insufficient lateral mobility to support multivalent binding. Limitations of this type of assay have been demonstrated previously (16). Thus, our observation that CD48 mAbs completely block 2B4-CD4d3+4 binding to cells does not rule out the existence of another ligand for 2B4. Ianelli et al. have reported that an hCD48–IgM fusion protein binds to epithelial cells, which do not express CD2 (31). It is unclear whether 2B4 is the CD48 ligand on these cells.

Our finding that the m2B4–mCD48 interaction in the mouse has a significantly higher affinity than the mouse or rat CD2–CD48 interactions (26, 27; Table 1) suggests that it is biologically significant. Moreover, the strength of the 2B4–CD48 interaction is conserved across species, with similar affinities being observed for the mouse and human 2B4–CD48 interactions (Table 1). Although fast, the 2B4– CD48 dissociation rate constants are comparable with values measured for other cell–cell recognition molecules, including the CD2–CD48 and CD2–CD58 interactions (27, 28). An interaction has been reported between hCD48 and hCD2 (32, 33), but this appears to be exceptionally weak (K_d >0.5 mM [28]). Since CD2 binds with a much higher affinity to CD58 and essentially all cells that express CD48 also express CD58, it is doubtful that the CD2–CD48 interaction is of physiological significance in humans. In contrast, the relatively high affinity of h2B4 for hCD48 suggests that this latter interaction is physiologically important.

The same CD48 mAbs block CD2 and 2B4 binding to CD48, suggesting that the 2B4 ligand binding site, like the CD2 binding site (34), lies on the GFCC′C′′ face of the V domain. Thus, a similar topology for the CD2–CD48 and 2B4–CD48 complexes can be predicted. The dimensions of the CD2–CD48 complex are predicted to be similar to the dimensions of the TCR–MHC complex (14 nm from membrane to membrane [7]). This has led to the suggestion that the maintenance of this intercellular distance is important for initiation of TCR–peptide MHC ligation (7). Interestingly, structure determination of an NK killer inhibitory receptor (KIR [35]) predicts that when bound to MHC class I, the complex will also span a similar distance. Together with functional studies (for a review, see reference 36), this suggests that CD2 on NK cells may contribute to activation through NK receptors. This hypothesis is supported by studies with mAbs and CD58-transfected cells that suggest that CD2 ligation can enhance activation of NK cells (36, 37). By analogy with CD2 (7) and substantiated by experiments with 2B4 mAb (10, 12), the 2B4–CD48 interaction may enhance NK and TCR ligand interaction as well as contributing to signals initiated through NK receptors and the TCR.

The expression on the same lymphocytes of two proteins (2B4 and CD2) that bind the same ligand (CD48) raises the question of whether CD2 and 2B4 have distinct functions. An analogy is exemplified by the well-characterized interactions of the T cell molecules CD28 and CD152, which bind the same ligands, CD80 and CD86 (38). There are several interesting parallels that can be drawn between these two pairs of molecules (CD28/CD152 and 2B4/CD2) that share ligands. First, they have different affinities for the same ligand, i.e., 2B4 and CD2 have different affinities for CD48, and CD152 has a 10-fold higher affinity than CD28 for CD80 (39). Second, their patterns of expression differ. The expression of both 2B4 and CD152 is much enhanced on activated cells (10, 11, 38), whereas both CD2 and CD28 are present on resting and activated cells. Third, they have different cytoplasmic regions. In the case of CD28 and CD152, this results in completely different functions, with CD28 stimulating and CD152 inhibiting (38). Thus, it seems likely that the cytoplasmic portions of 2B4 and CD2 will confer different functions. The more restricted expression of 2B4 on cells associated with non-MHC–restricted killing compared with CD2, which is present on virtually all T cells, suggests the primary role of 2B4 may be in generation of lytic activity. This is supported by the absence of effects of CD48 mAbs on target recognition (29).

Experiments have been done in vitro and in vivo with CD48 and CD2 mAbs that can now be reinterpreted in the light of there being two ligands for mCD48. For example, a CD48 mAb (HM48-1) has been shown to inhibit development of non-MHC–restricted lytic activity, whereas a blocking CD2 mAb had no effect (29). One explanation for this is that it is the 2B4–CD48 and not the CD2–CD48 interaction that has a major role in the development of the lytic population. In vivo, the synergistic effects of blocking CD2 and CD48 mAbs on allograft survival may be explained by disruption of the 2B4–CD48 and the CD2– CD48 interactions (30). Clearly, further experiments are needed to clarify this point. The activating properties (10) and partial inhibition of the 2B4–CD48 interaction make the currently available 2B4 mAb unsuitable for such experiments. Finally, recent observations that there are differences in phenotype between CD48- (40) and CD2-deficient mice (41, 42) can now be investigated with reference to the existence of an alternative ligand for CD48.

We thank Martin Wild and Liz Davies for helpful discussions and assistance with BIAcore™ analysis and reagents; Rita Mitnacht-Kraus for assistance with vector construction; Ruth Goddard for assistance with transfections and, with Mike Puklavec, for general cell culture; Amarjit Bhomra for DNA sequencing; Gillian Griffiths for helpful discussions; Dr. Yagita for HM48-1 mAb; and Simon Davis, Nasim Mavaddat, and Don Mason for critical reading of the manuscript.

CD4d3+4

domains 3 and 4 of rat CD4

CD4L

rat CD4 leader

h

human

IgSF

immunoglobulin superfamily

m

mouse

r

rat

RU

response unit(s)