Residues on both faces of the first immunoglobulin fold contribute to homophilic binding sites of PECAM-1/CD31

CD31 (PECAM-1) is a member of the immunoglobulin superfamily whose extracellular domain is comprised of six immunoglobulin-like domains. It is widely expressed on endothelium, platelets, around 50% of lymphocytes, and cells of myeloid lineage. CD31 has been shown to be involved in interendothelial adhesion and leukocyte-endothelial interactions, particularly during transmigration. CD31-mediated adhesion is complex, because CD31 is capable of mediating both homophilic and multiple heterophilic adhesive interactions. Here we show that the NH2-terminal (membrane-distal) immunoglobulin domain of CD31 is necessary but not sufficient to support stable homophilic adhesion. Key residues forming the binding site within this domain have been identified by analysis of 26 single point mutations, representing the most systematic analysis of a fully homophilic interaction between immunoglobulin superfamily family members to date. This revealed five mutations that affect homophilic binding. Uniquely, the residues involved are exposed on both faces of the immunoglobulin fold, leading us to propose a novel mechanism for CD31 homophilic adhesion.

CD31 (also known as PECAM-1, platelet-endothelial cell adhesion molecule) is a heavily glycosylated transmembrane protein of approximately 130 kDa (1, 2). It is widely expressed on circulating platelets, vascular endothelium, myeloid cells, around 50% of circulating lymphocytes (primarily CD8+, CD45RA naive lymphocytes), and CD34+ hemopoietic progenitor cells in human bone marrow. Molecular cloning identified CD31 as a member of the immunoglobulin superfamily (3-5), whose extracellular domain of 574 amino acids is comprised of six immunoglobulin-like domains, each encoded by a single exon (6), a membrane-spanning hydrophobic region and a cytoplasmic tail of 118 amino acids. Unusually among adhesion molecules, the cytoplasmic tail is comprised of eight exons from which it is possible to derive a number of splice variants (6). These splice variants have been identified as being expressed in a developmentally specific manner (7) and also in regulating CD31-mediated adhesion (8).

There is now significant evidence from both in vivo andin vitro studies to show that CD31 is involved in the extravasation of monocytes and neutrophils at inflammatory sites (9-13). These studies indicate that CD31 is a potential target for therapeutic intervention in both acute and chronic inflammatory conditions.

CD31-mediated adhesion is complex, because in common with other members of the immunoglobulin superfamily, such as neural cell adhesion molecule (14) and L1 (15, 16), it is capable of binding both to itself (homophilic adhesion) and to non-CD31 ligands (heterophilic adhesion). A number of potential heterophilic ligands have been identified, including the integrin αvβ3 (17, 18), a 120-kDa ligand on T-cells involved in down-regulation of T-cell responses (19) and an as yet uncharacterized glycosaminoglycan decorated ligand on L-cells (20). In addition, it has also been shown that CD31 is capable of up-regulating both β1- and β2-mediated adhesion following homophilic engagement (22-24).

Given the complexity of CD31 interactions and the wide distribution of potential ligands, it has proved necessary to study CD31-mediated adhesion in the context of heterologous systems. To define the domain or domains responsible for mediating homophilic binding, we have previously used chimeric fusion proteins comprised of the NH2-terminal 1, 1–2, 1–3, 1–4, 1–5, and 1–6 Ig domains of CD31 fused to the Fc portion of human IgG1 to form a nested series of deletions (24). COS cells expressing full-length CD31 allowed to adhere to surfaces coated with this domain deletion series of CD31-Fc proteins showed that domains 5 and 6 were necessary to support homophilic adhesion but did not exclude the possibility that other more NH2-terminal domains are also required. To address the roles of domains 1 to 4 in homophilic adhesion, we assessed the ability of antibodies to block homophilic adhesion. It was found that antibodies mapping to domain 1–2 of CD31 were able to inhibit the binding of CD31(D1-D6)+COS to CD31(D1-D6)Fc. We subsequently proposed a model in which the NH2-terminal domains of CD31 on the surface of one cell bind to the membrane proximal domains of CD31 expressed on the surface of an apposing cell and vice versa, in a fully interdigitating anti-parallel mode of adhesion. This double reciprocal model of adhesion has also been proposed for homophilic interactions mediated by other members of the immunoglobulin superfamily, for example carcinoembryonic antigen (25), neural cell adhesion molecule (26), and L1 (27). In the present study we set out to test this model and to identify specific residues involved in mediating CD31 homophilic adhesion.