Structural Motifs for Recognition and Adhesion in Members of the Immunoglobulin Superfamily
Total Page:16
File Type:pdf, Size:1020Kb
Journal of Cell Science 107, 2065-2070 (1994) 2065 Printed in Great Britain © The Company of Biologists Limited 1994 COMMENTARY Structural motifs for recognition and adhesion in members of the immunoglobulin superfamily Claire L. Holness and David L. Simmons* Cell Adhesion Laboratory, Imperial Cancer Research Fund, Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK *Author for correspondence INTRODUCTION The Ig domain found in recognition/adhesion molecules conforms to the β barrel plan. Compared with antibody V The immunoglobulin superfamily is the most abundant family domains, adhesion domains seem to have longer β strands and of cell surface molecules, accounting for 50% of leukocyte shorter connecting loop regions. The antigen-combining site of surface glycoproteins. This evolutionary success story is antibodies consists of the CDR (complementarity determining thought to be due to the stability of the Ig domain, which is regions) 1, 2 and 3, which are largely made from connecting able to resist the harsh proteolytic and oxidative environment loops. In contrast, adhesion binding surfaces seem to be made of the extracellular world. By mutation and selection, the Ig from the β strands themselves, arranged in flat faces. This is, domain has evolved to serve many different functions of course, a simplification and exceptions exist for both cases; including: receptors for growth factors (CSF-1 receptor, PDGF for antibodies the precise shape of the framework region of β receptor, FGF receptors); receptors for the Fc region of Ig (IgG strands is key to the organisation of the CDR loop regions. The receptors; CD16, CD32, CD64; CD89 IgA receptor); and as CDRs are not completely autonomous units; witness the trials adhesion molecules, which now seems to be the function of the of those presently engaged in humanising antibodies and gen- majority (CD2/CD58, CD28 and CTLA4, which bind to B7 erating in vitro antibody repertoires. For adhesion molecules, and B70; CD4/class II, CD8/class I, CD31/CD31, CD50 there are exceptions to the dominance of β strand face inter- (ICAM-3)/LFA-1, CD54 (ICAM-1)/LFA-1, CD102 (ICAM- actions, as some key residues involved in adhesion binding 2)/LFA-1, CD106 (VCAM)/VLA-4, NCAM/NCAM. L1, sites have been found in loops. MAG, TAG-1, CEA) (Springer, 1990). The packing and orientation of domains in adhesion The ideas and conventions defining the domain organisation molecules seems to be remarkably uniform. The contiguous Ig of this family were established by Alan Williams and Neil domains are arranged as an extended rod, with only small Barclay in a series of papers in the late 80s (Williams and angles of pitch between adjacent domains. For example, there Barclay, 1988; Williams et al., 1989). The basic Ig domain is is a common packing arrangement of domains 1 and 2 of CD2 built from a tightly packed barrel of β strands, either 7 strands and CD4, which is quite different from that seen in antibodies. in C1 and C2 domains (arranged in 2 layers of 3 and 4 strands So far, no antibody-like pattern, with heavy chain V domain designated ABE and CC′FG, respectively) or 9 strands in V bent at 90° to the contiguous CH1 domain, has been found. domains (arranged in two layers of 4 and 5 designated ABDE Although adhesin domains exist as an extended array, large and CC′C′′FG, respectively). There have been several recent rotational displacements between adjacent domains have been structural studies on adhesion molecules of the IgSF, including seen in the solved crystal structures of both CD2 and CD4 CD2 (solution NMR, Driscoll et al., 1991; Withka et al., 1993; (Jones et al., 1992; Ryu et al., 1990) and in the proposed model and crystallography at 2.8 Å resolution, Jones et al., 1992), CD4 for the N-terminal domains of ICAM-1 (Berendt et al., 1992). domains 1-2 (crystallography at 2.6 Å resolution, Garrett et al., For all these molecules adjacent domains are rotated by 160- 1993; Ryu et al., 1990; Wang et al., 1990) and domains 3-4 180° relative to each other. (crystallography at 2.8 Å resolution; Brady et al., 1993); the However, in spite of the common gross organisation of IgSF single V domain of CD8 (crystallography at 2.6 Å resolution; adhesins, there is a surprising diversity of modes of interaction Leahy et al., 1992); class II (crystallography at 2.8 Å resolu- amongst individual members. Fortunately, however, they can tion; Brown et al., 1993). In addition there have been extensive be grouped into a small number of patterns: (1) heterotypic mutagenesis screens of Ig domains involved in adhesion binding mediated by the N-terminal Ig domain(s); (2) including: CD2, which binds CD58 (LFA-3) (Somoza et al., homotypic binding mediated by ‘internal’ domain(s) and 1993; Arulanandam et al., 1993); CD4, which binds class II requiring anti-parallel alignment of interdigitated molecules. (Moebius et al., 1993); ICAM-1, which bind LFA-1 (Staunton et al., 1990; Berendt et al., 1992); and VCAM-1, which binds VLA-4 (Osborn et al., 1994; Vonderheide et al., 1994). Key words: Ig superfamily, adhesion, binding site 2066 C. L. Holness and D. L. Simmons Although the focus of this Commentary will be IgSF Key to symbols in all Figs: members involved in adhesion/recognition, the Ig fold has also evolved to bind small ligands such as growth factors. Some of Basic immunoglobulin related domain the recent work mapping their binding sites, which tend to be discrete binding pockets in internal domains, will be discussed. Type III fibronectin related domain α1 and β1 peptide binding domains of class II HETEROTYPIC CAMs AND THE DOMINANCE OF D1 Binding sites and surfaces within domains The most detailed molecular picture of IgSF domain recogni- tion comes from heterotypic interactions involving leukocyte Tyrosine kinase domain CAMs; specifically CD2/CD58 CD4/class II and ICAM- 1/LFA-1. Combinations of mutagensis screens with modelling A and/or structural analysis have localised binding sites to par- CD2 CD2/CD58 ticular domains, faces and key residues. The results indicate that in these receptor/ligand pairs, the IgSF binding surface is D2 D1 localised to one face, the CFG face, and key residues are located within that face and at its edges (Fig. 1). A general model for how IgSF domains of adhesion molecules interact CFG face was proposed by Springer (1991) based on the NMR structure and mutagenesis studies of CD2, and the precedent of how the domains of antibody molecules pack together. ICAM-1 (CD54) For human CD2, residues clustered in the β strands C (K34 B D5 D4 D3 D2 D1 and E36) and C′ (R48 and K49) and also charged residues in AAA the connecting loop between the F and G strands (K91 and AAA N92) have been shown to be critical in this interaction (Somoza AAA et al., 1993; Arulanandam et al., 1993). Similarly located AAA charged residues have been identified as key components of Mac-1 LFA-1 the binding site on ICAMs 1 and 3 for the leukocyte integrin LFA-1. For human ICAM-1 and ICAM-3, residues E37 (E34, ICAM-3 (CD50) ICAM-1) in the CC′ strand and Q75 (Q72, ICAM-1) within D5 D4 D3 D2 D1 domain 1 of ICAMs-1 and 3 are essential components of the LFA-1 binding site (Staunton et al., 1990; Fawcett et al., 1992; Vazeux et al., 1992; de Fougerolles et al., 1993; Holness et al., unpublished data). In addition, Li et al. (1993) have defined a peptide from domain 1 of ICAM-2 (covering residues 21-42; LFA-1 Staunton et al., 1989a), which inhibits binding of ICAM-2 to β ′ LFA-1. This peptide encompasses strands B, C and C and Class II dimer thus confirms the importance of this region for all the ICAMs C as they interact with LFA-1. The key residues, E34 and Q73, are conserved across species and are present in chimpanzee CD4 AAAAAAAA ICAM-1 (Hammond and McClelland 1993), rat ICAM-1 (Kita AAAAAAAA et al., 1992), mouse ICAM-1 (Horley et al., 1989) and mouse AAAAAAAA ICAM-2 (Xu et al., 1992). Indeed, this motif may be generally important for the inter- action of integrins with members of the immunoglobulin super- family (IgSF). To date, there are three examples of Class II dimer IgSF/integrin interactions: ICAMs with LFA-1 and/or Mac-1 (Simmons et al., 1988; Staunton et al., 1988, 1989a; Diamond Fig. 1. Binding sites in heterotypic CAMs. (A) CD2, and the interaction of CD2 with CD58; (B) ICAMs 1 and 3; (C) CD4 et al., 1991); VCAM with VLA-4 (α4β1) and α4β7 integrins (Osborn et al., 1992; Elices et al., 1990; Ruegg et al., 1992); interacting with a dimer of MHC class II molecules. and MadCAM-1 with the α4β7 integrin (Berlin et al., 1993). A tripeptide motif LDV containing a charged residue (D) at an ICAM-1 can also bind to the major group of human rhi- equivalent location on the adhesion domains of VCAM-1 (D1 noviruses and to an unidentified receptor expressed on the and D4) (i.e. within the C-C′ loop region) is also important for surface of Plasmodium falciparum-infected erythrocytes. In binding to its cognate integrin VLA-4 (α4β1), (Osborn et al., both cases the interaction surfaces are distinctly different from 1994; Vonderheide et al., 1994; and John Clements, British Bio- the LFA-1 binding face. ICAM-1 is thought to fit into a deep technology, personal communication). Thus, a structural motif canyon on the surface of the rhinovirus and the binding site consisting of two elements, an acidic residue (either D or E) encompasses a large area of the ICAM (Staunton et al., 1989b, located in a CC′ loop and additional residues in the FG loop 1990).