A Ligand for CD5 Is CD5

CD5 and CD6 are two unusual T cell surface receptors in that their extracellular regions are each composed of three scavenger receptor cysteine-rich (SRCR) domains instead of the more common Ig superfamily (IgSF) domains (1). SRCR domains are a primitive recognition domain stretching back to invertebrates and have been retained right through evolution in various tissues and the mammalian innate and adaptive immune systems (2).

Expression of CD5 (3–5), like CD6 (6), has an inhibitory effect on the cell that is not necessarily dependent on the extracellular region (3). High levels of CD5 are associated with resistance to activation (7–9). However, there is evidence of positive effects of CD5, like CD6 (6), that depend on extracellular engagement of CD5 and by logical extrapolation that there is an endogenous ligand. A soluble CD5-Ig fusion protein produced in vivo was inhibitory in a mouse experimental allergic encephalomyelitis (EAE) model (10). The effect was species specific, mouse but not human (h)CD5-Ig being inhibitory (10). Positive effects of CD5 were revealed in vivo in the mouse EAE model in which a cytoplasmic mutation in CD5 decreased the severity of disease (11). An overall role for costimulation by CD5 was shown by testing the effect of an immune challenge in the absence of CD5. Mice lacking CD5 were resistant to EAE (10). CD5, like CD6, is required for an optimal immune response (6, 12). The key to understanding the function of CD5 in the immune response is to define when and how the extracellular region is engaged and how this influences the signaling pathway regulated by CD5.

The heterophilic interaction between the membrane-proximal SRCR domain of CD6 and its IgSF ligand, CD166 (1), is not a general paradigm for the type of domain recognized by SRCR domains. Other types of recognition by SRCR domains include the interaction between another immune system receptor, CD163, and hemoglobin-haptoglobin complexes (13), and the recognition of pathogens (2, 14). The interaction between CD6 and CD166 (1, 12) is the only example that is understood in terms of a well-defined molecular interaction at the cell surface, of regulation of adaptive immune responses by a SRCR domain (6). Engagement of CD6 by CD166 in an Ag-specific immune response results in phosphorylation of CD6; direct recruitment of the adaptor protein, SLP-76, by the C-terminal tyrosine motif; and a net positive outcome (6).

Cell binding to immune cells has been observed with soluble forms of the extracellular region of CD5 (15–18) and with CD5 purified from cells (19). Consistent with it having a ligand-binding function, cell binding was affected by domain 1 alone (17). In none of these experiments was it possible to determine whether binding was mediated by biologically active material. CD5 mAbs did not block binding (17) consistently (19, 20). The cell surface binding partner for CD5 purified from cells was identified as CD72 (19). An interaction between CD72 and CD5 has not been reproduced. Soluble recombinant CD5 did not bind cell surface CD72 (15–17), and there was no evidence for productive engagement of cell surface-expressed CD5 and CD72 (16). Attempts to define the molecular identities of alternative cell surface binding partners for recombinant soluble CD5 have not been definitive (15–1ot8) probably due to its unusual lability, which may be related to a high degree of conformational flexibility, as shown by nuclear magnetic resonance (NMR) spectroscopy of CD5 domain 1 (21). We show that preservation of three-dimensional structure is key to molecular characterization of the ligand-binding properties of CD5, and the extensive functional data can be interpreted by homophilic interactions of CD5.

Chimeric proteins of the extracellular region of human (h; Genbank accession X04391) and mutants L26K, M124Q, V88D/V97K, E72K, L96P, and K69D (21); rat (r; Genbank accession X78985: bp59A → T, bp258 T → C); and mouse (m; Genbank accession D10728) CD5 were made. XbaI sites were introduced upstream of the initiation ATG (tctAGAAGGCCA), and SalI sites at the end of the extracellular region (join: gtg gca tcg tcg acc → VAS ST). A BamHI site (bp 2035) in hCD5 was removed. CD5, hCD6, and control CD4 leader constructs in frame with rCD4 domains 3 and 4 and a C-terminal peptide that can be biotinylated enzymatically (rCD4d3+3bio) in the pEFBOS-XB vector were expressed in 293T cells, supernatants biotinylated as described and tested for binding to a streptavidin-coated CM5 BIAcore chip (see below) (22). As the biotin tag reduced expression levels, the original rCD4d3+4 tag was used to construct hCD5CD4 and rCD5CD4 pEE14 vectors for production of stable lines (23). hCD5-His and rCD5-His pEE14 vectors were constructed by shuttling fragments via pEFBOS-XB containing a SalI site and His tag (STHHHHHH). CD4hCD72 chimera was made as a type II chimera with the C-terminal region of hCD72 containing the C-type lectin domain fused to the C terminus of the leader and domains 3 and 4 of rCD4. The amino acid sequence at the join was KGLNQQTNRV (CD4 in italics).

Stable lines were produced by transfection using Fugene into Chinese hamster ovary (CHO) K1 cells and clones identified by either inhibition ELISA or binding to mAb at 25°C immobilized via anti-mouse Fc on a CM5 BIAcore chip. Supernatant from typically one or two cell factories (Nunc, Naperville, IL) per selected clone was either purified by affinity chromatography on OX68 (rCD4d3+4 mAb)-Sepharose (Pharmacia, Peapack, NJ) with elution in 0.1 M Gly-HCl (pH 2.5) or nickel coupled to fast flow Sepharose with elution in 250 mM imidazole (pH 8). Eluted material was immediately dialyzed into HEPES-buffered saline or PBS. Eluate was then concentrated with a Viviflow 200 (10,000 molecular mass cutoff; Viviscience, Hannover, Germany) and subjected to gel filtration using Superdex 200 or 75 and fast flow liquid chromatography (Pharmacia). Monomeric and dimeric fractions were collected and analyzed by SDS-PAGE (nuPAGE; Invitrogen, Carlsbad, CA) under nonreducing and reducing conditions. Yields of purified protein were in the order ∼15 mg/l for His-tagged proteins and CD4 chimaeras. Theoretical molecular masses and molar extinction coefficients are as follows: hCD5CD4, 59,932 kDa, 75,390 M⁻¹ cm⁻¹; rCD5CD4, 60,137 kDa, 64,020 M⁻¹ cm⁻¹; hCD5-His, 40,445 kDa, 55,760 M⁻¹ cm⁻¹; rCD5-His, 40,650 kDa, 43,670 M⁻¹ cm⁻¹; mCD5-His, 39,762 kDa, 43,670 M⁻¹ cm⁻¹.

mAbs used as tissue culture supernatant or purified IgG were specific for the following: hCD5 domain 1, LEU1 (mIgG2a; American Type Culture Collection, Manassas, VA) and UCHT2 (mIgG1; gift of D. Cantrell, Dundee, U.K.); hCD5 domain 3, CD5-48 (mIgG1; produced against soluble purified hCD5CD4 by G. Roncador and J. Cordell, Oxford, U.K.); rCD5 domain 1, OX19 (mIgG1) (24); mCD5 domain 1, 53.7.131 (rIgG2a; American Type Culture Collection), rTCR mAb, R73 (24), hCD72, BU40; hCD6, T12 (American Type Culture Collection); rCD4 domains 3 and 4, OX68 (23); and control OX21 (IgG1). Recognition by CD5-48 mAb was mapped using hCD5CD4, hCD5 domain 1 CD4, rCD5CD4, and definitively with CD5 domain 3 made in Escherichia coli by A. Garcia-Garza, Mill Hill, London, U.K. (21). Fab fragments were prepared by papain digestion and gel filtration.

For analysis of homophilic interactions, CM5 chips were coated with streptavidin at 25°C, and then the temperature was raised to 37°C for immobilization of biotinylated CD4d3+4 chimeric proteins (5 μl/min) at equivalent levels, in the range of 1500–3000 response units (RUs) and subsequent injection (5 μl) of increasing concentrations of soluble protein (20 μl/min) to measure equilibrium binding (12, 22, 25). The rack containing proteins to be injected was cooled to 5–10°C to minimize aggregation and denaturation during the time scale of the experiment. Specific equilibrium binding was determined by subtracting responses in a control flow cell. mAbs were injected as tissue culture supernatant or purified (∼1–10 μg/ml) at 5 μl/min over chimeric proteins or polyclonal rabbit anti-mIgG (Pharmacia). CD5-His (∼1 μM) was injected (5 μl/min) over immobilized mAb. For analysis of mAb-blocking flow cells, double volumes of CD5-His were prepared and injected before and after saturation of immobilized CD5 with mAb. CD5 domain 1 mutants were immobilized on the chip from tissue culture supernatant via the rCD4d3+4 mAb, OX68, for screening their effect on homophilic binding by hCD5-His or hrCD5-His at 25°C.

Ag-specific IL-2 production of the mouse 2B4 hybridoma cells (10⁵) in response to moth cytochrome c peptide presented by CHO cells expressing I-E^K (CHO I-E^K; 5 × 10⁴–10⁵) was carried out as described previously in round-bottomed wells of 96-well plates (26). IL-2 production after 18–24 h was measured by ELISA. mAbs and Fab fragments were used at a final concentration of 5 μg/ml. Stimulation of lymph node cells from female Lewis rats was carried out as described (24). Following red cell lysis, lymph node cells were plated out at 2.5 × 10⁵/well in round-bottomed wells of 96-well plates and preincubated for 30 min at 0°C with Ab or Fab fragments at a final concentration, 5 μg/ml. Con A (5 μg/ml) and TCR mAb (R73) (5 μg/ml) were added and incubated along with the cells for 3 d at 37°C. Eighteen hours prior to harvesting, 0.5 μl [methyl-³H]thymidine (185 MBq/5 ml; Amersham, GE Healthcare UK, Little Chalfont, Buckinghamshire, U.K.) was added per well and incorporation into cells was determined by scintillation counting. Student t test was carried out on samples in triplicate, and significance was determined by p < 0.05.

Similarities in the composition of the extracellular regions of CD5 and CD6 suggested ligand-binding properties in common. However, during screens using surface plasmon resonance (SPR) for possible interactions of CD5 with cell surface proteins, including CD166 and the related protein named BCAM or Lutheran, there was no binding (data not shown). Instead, we noticed a weak self association of the purified recombinant hCD5CD4, a fusion protein composed of the extracellular region of hCD5 and domains 3 and 4 of rCD4. We examined the potential for CD5 for self association and, to eliminate the risk of denaturation due to purification at pH 2.5 using affinity chromatography, we produced His-tagged versions of hCD5 and rCD5 that were purified at neutral pH. On gel filtration, hCD5-His and rCD5-His separated as two peaks (data not shown), with elution volumes consistent with asymmetric proteins forming dimers. The majority of material was monomeric. Both peaks migrated as monomers on SDS-PAGE under nonreducing and reducing conditions consistent with a noncovalent association (Fig. 1A, 1B). Species-specific saturable binding by purified hCD5-His to immobilized hCD5 domain 1 (LEU1 and UCHT2) and domain 3 mAbs was observed, indicating that the material was antigenically active (Fig. 1C and data not shown). rCD5-His bound the rCD5 mAb, OX19 (data not shown). The hCD5-His gave some binding to the high levels of immobilized rabbit IgG (Fig. 1C). Such a slow on rate indicates this is probably due to traces of aggregated material.

We tested soluble CD5-His for self association by SPR and, to mimic physiological conditions, all experiments were carried out at 37°C. Biotinylated CD5CD4 fusion proteins that had not been subjected to denaturing purification conditions were immobilized on a streptavidin-coated chip and serial dilutions of hCD5-His injected (Fig. 1D). Specific homophilic interactions of hCD5-His were detected (Fig. 1D). The interaction was weak, as indicated by a rapid dissociation whether the dimeric or monomeric fraction was used (Fig. 1D and data not shown). Plotting the equilibrium-binding data for binding a series of CD5 concentrations revealed concentration-dependent species-specific binding of hCD5-His to hCD5CD4 (Fig. 1E).

In the complementary experiment, rCD5-His bound specifically to immobilized rCD5CD4 and mCD5CD4 equally well (see Fig. 2B and data not shown) and vice versa mCD5-His to rCD5CD4 and mCD5CD4 (see Fig. 2D, 2F). Thus, CD5 binds CD5 reproducibly in humans, mice, and rats. hCD5 reacts only with hCD5, whereas mCD5 and rCD5 can cross-react.

The behavior of soluble hCD5CD4 purified at pH 2.5 was analyzed in the same way as hCD5-His in the same experiment using the same chip (Fig. 1E, 1F). Soluble hCD5CD4 purified at pH 2.5 gave binding to both hCD5 and mCD5, but also the control protein, hCD6CD4, and showed none of the discrimination seen with hCD5-His.

The N-terminal domain of CD5 is topologically suited to mediate ligand binding (2). To determine whether species-specific homophilic binding depends on the N-terminal membrane distal domain of CD5, we constructed chimeric CD5CD4 proteins containing human domain 1 and rat domains 2 and 3 (hrCD5CD4; see Fig. 2E) and the complementary protein containing rat domain 1 and human domains 2 and 3 (rhCD5CD4) (Fig. 2). Soluble hCD5-His or rCD5-His was injected over equivalent levels of CD5CD4 or the chimeras (Fig. 2). Species-specific binding was calculated by subtracting responses in the rCD5CD4 flow cell for hCD5-His and vice versa for rCD5-His. hCD5-His bound equally well to hCD5CD4 and hrCD5CD4, showing that hCD5 domain 1 was necessary for species-specific homophilic binding (Fig. 2A). rCD5-His bound in a species-specific manner to rCD5CD4, but poorly to rhCD5CD4 (Fig. 2B). Nevertheless, the weak binding indicates that rCD5 domain 1 mediates species-specific homophilic binding. The identity and levels of immobilization of all four proteins were checked using mAbs specific for hCD5 domains 1 (LEU1) and 3 (CD5-48) and rCD5 domain 1 (OX19) (data not shown). rCD5 mAb, OX19, was shown to recognize domain 1 using a chimeric CD4 protein containing rat domain 1 and mouse domains 2 and 3 (data not shown).

Binding between hCD5-His and hrCD5CD4 shows that CD5 domain 1 is necessary for homophilic interactions. There are several possible topologies of homophilic interactions of CD5 that comply with this result. Theoretically, any one of the domains in hCD5-His could bind domain 1 of the immobilized material. To ask whether only CD5 domain 1 is involved in the species-specific homophilic interactions, we produced soluble hrCD5-His (Fig. 1B) and tested it for binding to hrCD5CD4. hrCD5-His bound equally well to hCD5CD4 and hrCD5CD4, but not to rCD5CD4 or rhCD5CD4 (Fig. 2C). These data show that hCD5 domain 1 binding to itself is the dominant interaction in the species-specific homophilic interactions. In the complementary experiment with the same chip (Fig. 2D), mCD5-His bound the chimeras containing rCD5 domain 1, confirming the identity and activity of the chimeras on the chip and the cross-reactivity between mCD5 and rCD5 homophilic interactions. The validity of CD5 homophilic binding is emphasized by it being demonstrated in three species and with two types of recombinant proteins, His-tagged and CD4d3+4 fusion proteins.

An understanding of the potential of CD5 domain 1 mAbs to block ligand binding by CD5 aids in mapping the ligand binding site on CD5 domain 1 and is important for interpretation of their effects in functional assays. Two hCD5 domain 1 mAbs, LEU1 and UCHT2, recognize epitopes that do not overlap (21). We analyzed the effect of saturating immobilized hCD5CD4 with LEU1 on homophilic binding by hCD5-His and hrCD5-His. LEU1 specifically blocked binding of hCD5-His to hCD5CD4 (data not shown and Fig. 2C). Blocking of homophilic binding by UCHT2 was not observed (see Fig. 3B).

We tested the rCD5 domain 1 mAb (OX19) for its effect on CD5 ligand binding. OX19 blocked rCD5 homophilic binding (Fig. 2D). Similarly, saturation of mCD5CD4 with a mAb (53.7) specific for domain 1 (27) inhibited mCD5 homophilic binding (Fig. 2F). Cross species blocking by the rCD5 (OX19) (Fig. 2D) and mCD5 (53.7) mAbs (data not shown) showed that soluble CD5-His did not compete for immobilized mAb.

A clear separation between the epitopes of LEU1 and UCHT2 was demonstrated by a single mutation, L26K, in hCD5CD4, which specifically abolished recognition by UCHT2 mapping its epitope to the N terminus of CD5 domain (21). The L26K hCD5CD4 mutant, which is not recognized by UCHT2, bound CD5-His as well as wild-type CD5 (Fig. 3A) consistent with the N terminus of CD5 domain 1 not being essential for homophilic interactions. The mutant gave a clearer result than testing for blocking with UCHT2, which appeared to cross-link immobilized and soluble CD5 whether it be the wild-type or a nonbinding mutant (Fig. 3B). Mutation of a residue, M124Q, in the region defined as critical for the interaction between CD6 and CD166 did not have a drastic effect on homophilic binding (Fig. 3A). We analyzed a panel of hCD5 domain 1 mutants for their effect on homophilic binding (Table I). Two mutants showed reduced binding, E72K and V88D/V97K (Fig. 3B, Table I). Expression and LEU1 mAb binding of E72K were reduced relative to wild-type CD5, which may indicate structural instability. V88D/V97K identified as nonbinding is the mutant from which the NMR structure (21) was derived and thus maintains structural integrity. In the screen of the panel of mutants, immobilized recombinant hCD72 was used as a negative control and its identity checked with the CD72 mAb, BU40.

We developed a well-controlled in vitro assay in which we could detect functional effects of engagement of CD5 domain 1. Wild-type and mutant (V88D/V97K) CD5 were expressed in 2B4 mouse T hybridoma cells. The hCD5 variants were highly expressed as monitored by CD5 domain 1 and 3 mAbs (Fig. 4A). Consistently, CD3 levels were higher on cells expressing mutant CD5 (Fig. 4A). Endogenous mCD5 was downregulated on cells expressing hCD5, whereas transfected mCD2 remained unchanged (Fig. 4A). The cells transfected with mCD2 were used as the parental line to boost levels of Ag-specific IL-2 in the stimulation assays. Cells were stimulated with peptide presented by CHO cells expressing the MHC class II, I-E^K (CHO I-E^K cells), or CHO I-E^K cells expressing mCD48 (28). A similar pattern of responses was observed in response to both types of APC with higher IL-2 production for the ones expressing mCD48 (data not shown).

Ag-specific IL-2 production by cells expressing mutant CD5 was increased relative to wild-type CD5 (Fig. 4B). To ensure this was related to engagement of hCD5 domain 1 and not just the higher CD3 levels on the mutant cells, the effect of a soluble blocking CD5 mAb on responses by cells expressing wild-type hCD5 was analyzed. The hCD5 mAb that blocked species-specific homophilic interactions enhanced Ag-specific IL-2 production (Fig. 4C). Confirmation that this was due to blocking was obtained by showing a Fab fragment also enhanced IL-2 production (Fig. 4C). Functional effects of hCD5 domain 1 in mouse cells are consistent with species-specific homophilic interactions, resulting in productive engagement.

Dissection in the hybridoma model revealed the potential of the extracellular region of CD5 to regulate inhibitory effects of CD5. To assess its physiological relevance, we examined the potential of the extracellular region of CD5 to regulate inhibition of immune responses by a normal polyclonal cell population. We analyzed the mode of action of the rCD5 mAb, OX19, which has been shown to have costimulatory effects on primary cells (24). The rCD5 mAb, OX19, blocks homophilic interactions (Fig. 2C). Comparison of the effects of freshly gel-filtrated Fab fragments with Fab′₂ or IgG [data not shown (24)] showed that all three reagents enhanced polyclonal proliferation of rat lymph node cells (Fig. 4D). CD5 was not modulated off the surface by OX19 IgG or Fab in vitro (data not shown). The data are consistent with homophilic interactions of CD5 regulating activation of normal cells.

Molecular characterization of homophilic binding by CD5 has resolved two major sources of confusion over immune regulation by this receptor. First, specific and nonspecific binding by the extracellular region of CD5 were distinguished. CD5 interacted with itself in a species-specific manner. Second, an understanding that the extracellular region of CD5 can engage in homophilic binding provides a molecular mechanism for functional effects of perturbing engagement of the extracellular region of CD5.

The biochemical and functional data indicated that homophilic CD5 interactions were strong enough to mediate productive interactions to regulate immune responses. The micromolar concentrations of soluble protein that yielded measurable binding to immobilized material were in the range of those observed for other weak, but physiological relevant interactions (12, 25). The kinetics of CD5 binding revealed rapid on and off rates consistent with transient interactions typical for cell surface proteins (12, 22). Low stoichiometry precluded good mathematical analysis, as we have done for other homophilic interactions (12, 25). Low stoichiometry of binding measured by SPR could be due to the presence of inactive protein or a consequence of self association by the soluble material competing with immobilized.

The presence of a region of particular flexibility in the NMR solution structure of hCD5 domain 1 (21) provides a molecular basis for the difficulty in identifying specific ligand binding. Interactions of CD5 across species (15, 16, 19) or on cells that do not express CD5 (17) (A. Johnstone, unpublished observations) may be explained by the capacity of denatured CD5 in CD5 preparations to give nonspecific binding. Promiscuous binding by denatured CD5 may reflect a general propensity of SRCR domains, including CD5, to multerimize (14, 29, 30).

Homophilic binding by CD5 domain 1 may provide more of a general paradigm for recognition by SRCR domains than the interaction between an IgSF domain of CD166 and the membrane-proximal SRCR domain of CD6. Domain 1 of CD6 does not participate in the interaction with CD166 (1, 12), but by analogy with CD5, its membrane distal position indicates a ligand-binding function. There are data to indicate a significant role for CD6 domain 1 in regulation by CD6 in immune responses. Inhibitory effects of two CD6 mAbs characterized as specific for different epitopes on the membrane distal N-terminal domain have been described (31–34). As they were effective in soluble form, it is likely they are blocking an interaction. Importantly, the T1 mAb has shown therapeutic immunosuppressive effects in patients (33, 34). This could involve a homophilic interaction of CD6 domain 1. A ligand-binding function of CD6 domain 1 is also suggested by the existence of a form of CD6 that lacks the membrane-proximal CD166 binding domain on activated T cells (35).

Homophilic binding by CD5 domain 1 has the potential to mediate interactions between cells, in trans or on the same cell, in cis. The lack of involvement or of a subsidiary role for domains 2 and 3 in CD5 homophilic interactions favors trans interactions as might occur between T cells or T and B cells (16). Alternatively, the absence of CD5 on dendritic cells at the initiation of an immune response favors cis interactions. The two topologies for CD5 homophilic binding are not mutually exclusive. There is good precedent for competition between cis and trans interactions in the Siglec family (36) and interactions between Ly49 and MHC class I (37). The highly conserved flexible stalk between CD5 domains 1 and 2 is likely to be important in regulating potential trans versus cis CD5 interactions, as has been recently shown for Ly49 (37). The exclusive role of CD5 domain 1 in species-specific homophilic interactions differs from the involvement of all three SRCR domains of CD5 in an interaction with fungal wall components (38). The involvement of domains 2 and 3 of CD5 in other interactions (39–41) is not excluded by the homophilic binding data.

Correlation of activation and inhibition by CD5 with a requirement for extracellular engagement is not absolute. Dissection of engagement of CD5 in the cell line model shows the potential of the extracellular region of CD5 to regulate its inhibitory effects. Caution must be exercised in dismissing the extracellular region in regulating inhibition by CD5 based on experiments with genetically manipulated mice (3). To demonstrate a role for the extracellular region of CD5, it is important to distinguish between engagement of the extracellular and intracellular region. Simply observing differences between CD5⁺ and CD5⁻ cells does not make this distinction. An inhibitory role for the cytoplasmic region of CD5 was shown in mice expressing CD5 lacking an extracellular region (3). To test whether interactions of the extracellular region contribute to regulation by CD5, we used a CD5 mAb that blocks CD5 homophilic interactions. In retrospective analysis of the literature with the knowledge that a particular CD5 mAb blocks homophilic binding, rigorous analysis of the mode of action is required before functional effects of the mAb, for example, OX19 (24), can be interpreted. To ensure the mode of action of OX19 was blocking, we used a Fab fragment. Enhancement of proliferation by rat lymph node cells by Fab fragments from OX19 provides good evidence that the extracellular region of CD5 is important for controlling activation in normal immune responses. A molecular understanding of the effect of ligand binding by the extracellular region of CD5 allows the relative contributions of extracellular and intracellular engagement of CD5 to be more clearly delineated and provides a molecular basis for interpreting functional data.

Homophilic binding by CD5 provides a molecular basis for species-specific inhibition of EAE in mouse by soluble CD5-Ig (10). The mCD5 mAb (53.7) that blocks homophilic interactions has also been reported to have functional effects, inhibiting B cell activation (16), which, depending on the mode of action of the mAb in these experiments, would be evidence for ligand-dependent positive effects of CD5 in T-B cell collaboration. The difference between the overall effect of CD5 engagement being activating or inhibitory may depend on the number and activation status of responding cells in a polyclonal population. The evidence is that the extracellular region of CD5 is important for regulating both activation and inhibition by CD5 and the molecular mechanism involves species-specific homophilic interactions of CD5 domain 1.

We thank the following for support in this project: Steve Simmonds, Andrew McGrath, Alan Johnstone, Nicholas Clarkson, members of the laboratory, and Neil Barclay.

Disclosures The authors have no financial conflicts of interest.