L1 Makes Immunological Progress by Expanding Its Relations

The cell-adhesion molecule L1 was originally described in the nervous system. It has recently been detected in CD4+ T lymphocytes, peripheral B lymphocytes, and granulocytes in the human immune system and in similar leucocyte types in the murine immune system. L mediates neural recognition by Ca+2, Mg+2-independent homophilic binding. In the human and murine immune systems, L1 binds to the “classical” vitronectin receptor, αVβ3, and fibronectin receptor, α5β1, respectively, and abstains from homophilic binding. Homophilic L1 binding probably involves antiparallel alignment of several interactive domains. Integrin binding is mediated by a short segment of immunoglobulinlike domain 6, which includes two RGD repeats in rodent L1 and one RGD motif in human L1. L1 is modulated in activated leucocytes in vitro in parallel to L-selectin, and diverse cell types release intact L in vivo and in vitro. Released L1 can bind to laminin and adheres to the extracellular matrix of sciatic nerve, M21 melanoma, and possibly spleen and other tissues. It can support integrin-dependent cell migration and preliminary data implicate it in tumor development and transnodal lymphocyte migration.


INTRODUCTION
The development of a functional cellular immune system and effective immune response depends on selective leucocyte trafficking to, from, and within organs of the immune system as well as to variable but specific addresses such as inflammatory sites.
leucocyte extravasation have produced ample information on the mechanisms mediating leucocyteendothelial recognition (for extensive reviews, see Butcher, 1991;Imhof and Dunon, 1995). There is, however, little information on the adhesion mechanisms that mediate leucocyte migration through and beyond the endothelial wall, through lymph glands and inflamed tissues.
During the last year, the cell-adhesion molecule L has emerged as an important candidate for mediating such processes. L1 is primarily expressed by neural and blood cells involved in targeted migration and is at present the only known cell-adhesion molecule that is abundant in both nervous and immune systems. Its importance in human ontogenesis is underlined by the severity of neurological disorders associated with mutations in the L1 gene. These include hydrocephalus and mental retardation and have recently been united under the acronym CRASH syndromes (Fransen et al., 1995). Associated manifestations in the immune system remain to be defined. In the nervous system, L1 is predominantly expressed by neuronal subsets and Schwann cells. It mediates neuronal adhesion, axonal navigation and fasciculation, axon-Schwann cell recognition, and possibly granule-cell migration. In the immune system, L1 is expressed by neutrophils and lymphocyte subsets (Table I). Its function(s) in the immune system is not certain, but preliminary studies using tumor lines, T and B lymphoblasts suggest that L1 may be involved in both leucocyte-leucocyte and leucocyte-endothelial cell interactions.

L1 LIGANDS
The functional vicissitude of L has been attributed to its recently discovered diversity of binding mechanisms (Table II). In rodents and man, L1 can bind to at least three counterreceptors with distinct cation requirements. The major L1 ligand for L1-expressing murine leucocytes and lymphoma ESb-MP cells appears to be the "classical" fibronectin receptor, ce5fll integrin (Ruppert et al., 1995

FIGURE
Schematic representation of L1. (A) L1 is a transmembranal glycoprotein with six immunoglobulinlike (Ig1-6) and five fibronectin type III-like (FNInl-5) domains in its extracellular segment. Two proteolytic sites (broken lines) generate characteristic fragments (vertical double arrows) in vivo and in vitro. Alternative splicing of two microexons generates several isoforms. In neurones, the intracellular tail can possibly be anchored to spectrin via a neural isoform of ankyrin. The positions of potential a5/31 (ETA, RGDG, RGDS, NGR), a4/31 (LDV), and FGF receptor-homology sequences (APY, APYW), basic box (+), cystein residues outside (*) and inside (S) Ig domains, and conserved tyrosine (Y) and tryptophan (W) residues in FNI domains are indicated. Clouded areas denote putative interactive segments mediating homophilic binding. (B). The glycan structure of HNK-1and L3-immunoreactive glycolipids. L1 variably expresses both epitopes. HNK-1and L3-immunoreactive glycolipids inhibit L1 binding to laminin and NCAM, respectively. HNK-l-immunoreactive glycolipids also bind to ,-and I-selectins.  (Ebeling et al., 1996) and melanoma M21 cells (Montgomery et al., 1996), L1 strongly binds to cV/33 and weakly to c5/31. The binding between L1 and oV/33 requires Ca +2 and Mg +2 but no prior activation of the cells, although Mn +2 may enhance the binding. In the nervous system, L1 mediates neuronal recognition by temperature-and cation-independent homophilic binding (Lemmon et al., 1989). It is at present not certain that neuronal L1 recognizes additional ligands. Binding between neuronal L1 and integrins has to our knowledge not been examined under conditions that support integrin-L interactions.
However, purified L1 from murine neuroblastoma N2A cells binds murine c5/31 and serves as a good substrate for c5/31-mediated leucocyte adhesion in the presence of Mn +2 in vitro (Ruppert et al., 1995). Purified L1 from rat pheochromocytoma PC12 cells (NILE) interacts with human cV/33 and possibly with o5/31 (Montgomery et al., 1996). There are at present no indications that L1 binds to mammalian homologues of the chicken neuronal adhesion molecules axonin-1 or Fll, the chondroitin sulphate proteoglycan neurocan, or the presumptive astrocytic ligand(s) that is recognized by the chicken homologue of L1, NgCAM (for a review, see Brtimmendorf and Rathjen, 1993).

BINDING MECHANISMS OF L1
Homophilic Binding Holm et al. (1995) have reported that microbeads coated with intact L1 or with fusion proteins derived from L1 domains Ig3-4 or FNnI3-5 strongly adhere to substrate-immobilized L1. Microbeads coated with fusion proteins derived from other short segments of L1 or from longer segments that include these domains underwent homotypic aggregation but weaker or no adhesion to substrate-immobilized L1.
These observations appear to indicate that mutual attraction among parallel domains of L1 could facilitate L1 clustering on the cell surface, but trans binding between L1 molecules may primarily involve domains Ig3-4 and FNnI2-4 ( Figure 2A). Systematic analysis of the homophilic binding mechanism of L1 is still wanting. However, the data are consistent with the proposed model for the homophilic binding mechanisms of carcinoembryonic antigen (CEA; Zhou et al., 1993) and platelet-endothelial cell-adhesion molecule (PECAM-1, CD3 l; Fawcett et al., 1995). Like L1, CEA and CD31 belong to the immunoglobulin supergene family and function as Ca+2-independent cell-adhesion molecules. Homophilic CEA binding involves double reciprocal interactions between the amino terminal and an internal domain. Homophilic CD31 binding is primarily mediated by domains 2, 3, 5, 6, but requires the presence of all six domains. It has therefore been suggested that stable homophilic binding is mediated by antiparallel alignment of apposing molecules and correct interdigitation of corresponding "high-affinity" domains ( Fig. 2A).
Heterophilic Binding to aVfl3 and a5/l lntegrins The major integrin-binding site of L1 lies in a segment of domain Ig6 that includes the consensus integrin-binding motif arginine-glycine-aspartate (RGD; Fig. 2B and Fig. 2C)  denoted by clouded shading. Adopted from Haas and Plow, 1994. from domain Ig6 of human L predominantly binds to cV/33 and supports the adhesion of cV/33-expressing MED-B 1 cells but not c5/31-expressing Nalm-6 cells (Ebeling et al., 1996). On the other hand, the synthetic peptide CWRGDGRDLQERGDSDK derived from domain Ig6 of murine L1 binds to o5/31, inhibits the homotypic aggregation of murine lymphoma ESb-MP cells, and supports c5/31-dependent adhesion of murine peripheral thymocytes (Ruppert et al., 1995). By amino acid sequence comparisons between domain Ig6 of L1 and immunoglobulinlike domains with known crystal structure, domain Ig6 of L1 most closely resembles domain Ig2 of VCAM-1. It potentially has a C2-set immunoglobulin fold (Fig. 2C) with a short C' strand and particularly elongated C'-E loop. Rodent and human L 1 contain a conserved RGD motif (RGDS; Fig. 2B) in the putative C-C' loop. A second RGD motif (RGDS; Fig. 2B) is expressed by rodent L1 in the putative C'-E loop. However, crystallographic analysis has not been performed, and it cannot be excluded that both RGD motifs are expressed on the same /3 loop (C-C' or C-D). In human L1, arginine to leucine change disrupts the RGDS motif in the C'-E loop, whereas a lysine to arginine change generates an "inverse" RGD (DGR) motif at a different pogition, presumably on the same /3 loop (Fig. 2C).

Mechanisms of Selective Integrin Binding
Since L1 is expressed by similar neuronal and leucocyte subsets in mouse and man, the existence of variable RGD sites in the otherwise highly conserved molecule may reflect evolutionary adaptation of L to binding functionally analogous integrins in different species. The tripeptidyl RGD is expressed by most integrin-binding matrix proteins and serves as the main binding site for most integrin-substrate adhesion molecules. However, different integrin dimers bind to different RGD-expressing proteins, indicating that the molecular context is a determinant factor in selective RGD recognition (for reviews, see Ruoslahti, 1988;Haas and Plow, 1994). RGD recognition by c1/31 and c2/31 depends on RGD presentation in a flexible conformation at the apex of a type II/3 loop with the basic and acidic side chains of arginine and aspartate pointing away from one another (Fig. 2D). A5/31 recognizes cyclic RGD-containing peptides, suggesting similar conformational constraints, whereas cV/33 binds both to cyclic and linear RGD-containing peptides (for a review, see Haas and Plow, 1994).
A5/31 strongly binds to murine (Ruppert et al., 1995) but not human (Ebeling et al., 1996) L1. Therefore, c5/31 possibly recognizes the fibronectinlike motif RGDS of rodent L1 (Fig. 2B) that is missing in human L1. AV/33 appears to bind to the conserved RGDG site of L1 (Fig. 2B), since M21 cell adhesion to human L1 is abolished by mutation of this motif (Montgomery et al., 1996), and human cV/33 mediates cell binding to human, rat (Montgomery et al., 1996), and murine L1. However, binding of ozV/33 to L1 has so far not been detected in murine cells. It is therefore likely that species-specific differences in the binding properties of oV/33 and possibly codetermine the choice of L ligand. It is at present an open issue whether amino acid changes to the carboxyl terminal to the RGD motifs of L1 (Fig. 2B) are also important to selective integrin-L1 binding. In addition to conformational constraints on RGD presentation, conjoint non-RGD binding sites in the ligand may also contribute to selective ligand recognition by RGD-binding integrins. Binding of cV integrin(s) to collagen is enhanced in the presence of short noncollagen RGD peptides, indicating that RGD recognition activates cryptic collagen-binding sites in cV integrins (Agrez et al., 1991). Binding of c5/31 to the RGD motif in domain FNn10 ("cell-binding domain") of fibronectin is imperative to stable contact between c5/31 and matrix-embedded fibronectin. Yet, the affinity of c5/31 to fibronectin is strongly reduced by mutations in domains FNm8 and FNm9 of fibronectin (Obara et al., 1988;Aota et al., 1991), suggesting that additional non-RGD integrin-recognition sites are also important. Interestingly, L1 shares with domains FNIn8 and FNm9 of fibronectin several non-RGD peptide sequences that have been implicated in c5/31 binding (Fig. 1). Preliminary studies from our laboratory indicate that the synthetic cycloheptapeptide *RRETAWA*, resembling one of those sites, interferes with Ll-c5/31 binding through action on a non-RGD site of L 1. However, additional studies will be required to determine the significance of this observation and whether L1 domains other than domain Ig6 indeed participate in regulating the integrin-binding potential of L1.

LI RELEASE AND INTERACTION WITH THE EXTRACELLULAR MATRIX
Possible Roles of Homophilic and Integrin Binding in Ll-dependent Adhesion, Migration, and Signal Transduction L1 joins a growing number of cellular receptors for "classical" matrix-binding integrins (for a review, see Imhof and Dunon, 1995). Among these, CD31 is of particular interest since L1 and CD31 appear to have similar homophilic binding mechanisms ( Fig. 2A) and high affinity for ceV/33 (Piali et al., 1995;Montgomery et al., 1996) but are expressed by mutually exclusive lymphocyte subsets (Ebeling et al., 1996).
The functional significance of these dual binding mechanisms is not clear. However, it seems possible that homophilic L1 binding primarily mediates cellcell recognition and stimulation of FGF receptormediated signal transduction, whereas L l-integrin binding may support cellular locomotion. Ll-transfected J558L cells that adhere to immobilized L1 by homophilic binding undergo very limited movement on substrate L 1. Conversely, L 1-expressing M21 cells interact with L1 via ceV/33 and exhibit pronounced ceV/33-dependent haptotactic movement on substrate L1 (Montgomery et al., 1996). In addition, monoclonal L1 antibody 324 that specifically inhibits L1integrin interactions does not perturb the short-term aggregation of murine cerebellar neurones and neuroblastoma N2A cells (Rathjen and Schachner, 1984). But it inhibits the migration of granule-cell neurones (Lindner et al., 1983) and neurite elongation (Fischer et al., 1986) in cerebellar explants from postnatal mice. Moreover, homophilic L1 binding activates FGF receptor-dependent signal transduction in rat neurones, but probably not nonreceptor-type tyrosine kinases (Doherty et al., 1995). However, neurite extension by murine cerebellar neurones on immobilized L1 is specifically abolished by homozygous src deletion (Ignelzi et al., 1994), and L cross-linking by Several types of neurones release increased amounts of L1 in response to nerve growth factor in vitro (Richter-Landsberg et al., 1984). Melanoma M21 cells spontaneously release L1 at an average rate of approximately 104 molecules/cell/hour (Montgomery et al., 1996), and lymphocytes downregulate L1 expression upon activation in vitro and in vivo (Hubbe et al., 1993). Stimulation of neutrophils with the phorbolester PMA leads to rapid loss of L1 from the cell surface with remarkably similar kinetics to the shedding of i-selectin. L-selectin shedding has recently been shown to depend on an MMPl-like metalloendoprotease (Preece et al., 1996). Whether a similar mechanism also applies to L1 remains to be examined. Released L1 can be detected in association with the extracellular matrix of murine sciatic nerve (Martini and Schachner, 1986), and within intratumor laminin strands of human melanoma M21 (Montgomery et al., 1986). Selective binding between purified L1 and laminin has been demonstrated (Hall et al., 1993), but the physiological function of matrixembedded L1 is still uncertain. Conceivably, matrixembedded L1 gradients may locally promote targeted cell migration as a substrate for integrin adhesion. Recent experiments using recombinant L1 have shown that L1 indeed can support integrin-mediated cell migration (Duzcmal et al., 1997). The possibility that released L1 may promote intratumor angiogenesis by interaction with ceV integrins of sprouting vessels can also be considered.

CONCLUSIONS
The cell-adhesion molecule L1 serves both as a celladhesion molecule and as a matrix-embedded substrate for cell migration. In both functions, L1 may undergo homophilic interaction or bind to "classical" integrin-matrix receptors. Dual homophilic and integ-. rin binding has also been reported for the nonneural recognition molecules CD31 and E-cadherin. Homophilic binding has primarily been associated with early recognition events. Binding between homophilic cell-adhesion molecules and integrins appears to stabilize cell connections and promote cell motility. This information implies that the major difference between cell-cell and cell-matrix interactions pertain to early recognition and choice of contact. But the mechanisms mediating long-term adhesion and migration on cellular and matrix substrates may be related.