Cis-activation of L1-mediated ankyrin recruitment by TAG-1 homophilic cell adhesion.

Neural cell adhesion molecules (CAMs) of the immunoglobulin (Ig) superfamily mediate not only cell aggregation but also growth cone guidance and neurite outgrowth. In this study we demonstrate that two neural CAMs, L1-CAM and TAG-1, induce the homophilic aggregation of Drosophila S2 cells but are unable to interact with each other when expressed on different cells (trans-interaction). However, immunoprecipitations from cells co-expressing L1-CAM and TAG-1 showed a strong cis-interaction between the two molecules in the plane of the plasma membrane. TAG-1 is linked to the membrane by a glycosylphosphatidylinositol (GPI) anchor and therefore is unable to directly interact with cytoplasmic proteins. In contrast, L1-CAM-mediated homophilic cell adhesion induces the selective recruitment of the membrane skeleton protein ankyrin to areas of cell contact. Immunolabeling experiments in which S2 cells expressing TAG-1 were mixed with cells co-expressing L1-CAM and TAG-1 demonstrated that the homophilic interaction between TAG-1 molecules results in the cis-activation of L1-CAM to bind ankyrin. This TAG-1-dependent recruitment of the membrane skeleton provides an example of how GPI-anchored CAMs are able to transduce signals to the cytoplasm. Furthermore, such interactions might ultimately result in the recruitment and the activation of other signaling molecules at sites of cell contacts.

The navigation of growth cones to their targets in the developing embryonic nervous system is a critical step in the patterning of neuronal projections. Substantial evidence suggests that axonal guidance depends in part on cell surface and on extracellular matrix molecules, which are expressed along the path of the advancing growth cones (1,2). Axonal extension and pathway choices are influenced by membrane receptors to these molecules, which are expressed on the growth cone.
One group of such molecules are cell adhesion molecules (CAMs) 1 belonging to the immunoglobulin (Ig) superfamily that are expressed by advancing growth cones and are able to recognize and transduce environmental signals (3). Ig domain CAMs have been implicated to act both as receptors as well as substrates for growing axons (4,5). For example, members of the L1 family of neural CAMs, such as mammalian L1-CAMs and chicken Ng-CAM, not only exhibit a strong Ca 2ϩ -independent homophilic adhesive activity (6,7), they also promote neurite outgrowth in culture, probably by the activation of neuronal FGF receptors (8). Through the interaction with the cytoplasmic linker protein ankyrin, L1 family members are also connected to the membrane skeleton (9). Human L1-CAM and the Drosophila L1 homologue neuroglian both recruit ankyrin and other components of the membrane skeleton to cell contact sites in Drosophila S2 cell aggregates (10,11). This interaction strictly depends on the extracellular L1 adhesive activity, and ankyrin binding in turn stabilizes the L1 adhesive interaction (12).
Members of another subgroup of Ig domain neural CAMs are anchored in the plasma membrane by a glycosylphosphatidylinositol (GPI) moiety. These include TAG-1 in mammalian species and its chicken homologue axonin-1 (13,14). TAG-1/ axonin-1 has been shown to mediate homophilic cell adhesion and is able to promote neurite outgrowth in culture (13,15,16). However, TAG-1-induced neurite growth is not only mediated by but also requires other neuronal membrane proteins, such as L1-or ␤ 1 integrin-type molecules (17,18). TAG-1/axonin-1 interacts with a number of different heterophilic binding partners, including several members of the L1 family, NCAM, nervous tissue-specific chondroitin sulfate proteoglycans, and several extracellular matrix molecules (18,19). It has been suggested that some of these heterophilic interactions might enable TAG-1 to induce or influence intracellular signaling processes (5). Axonin-1 and Ng-CAM expressed in the same plasma membrane engage in a strong cis-interaction, forming larger multimeric complexes that are also associated with several intracellular protein kinase systems (18,20). This cisinteraction between TAG-1/axonin-1 and L1-type molecules is essential for the stimulation of neurite outgrowth on TAG-1/ axonin-1 substrates in culture, and homophilic TAG-1 cell adhesion appears to activate TAG-1-associated L1-CAM molecules (16,17). Although several lines of evidence suggest that the interaction of TAG-1 with L1-CAM is an important link in TAG-1-initiated signal transduction, a direct demonstration that TAG-1 is capable of altering the functional state of L1-CAM has been missing. In our present study we demonstrate that TAG-1-mediated homophilic cell adhesion induces an intracellular restructuring of the membrane skeleton by the cisactivation of human L1-CAM.

EXPERIMENTAL PROCEDURES
Materials and Antibodies-A mouse polyclonal serum was raised against a glutathione S-transferase-Drosophila ankyrin fusion protein, which has been previously described (21), and used at a dilution of 1:200 for indirect immunofluorescence microscopy. Rabbit anti-L1-CAM was a gift from Dr. Vance Lemmon (Case Western Reserve University, Cleveland, OH), and the rabbit anti-TAG-1 serum was characterized previously (22,23). Fluorescein isothiocyanate-conjugated goat antimouse IgG, Texas Red-conjugated goat anti-rabbit IgG, and horseradish peroxidase-conjugated goat anti-rabbit IgG were purchased from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA). Schneider's medium, penicillin/streptomycin stock solution, and fetal calf serum were from Life Technologies, Inc.
Transfected Drosophila S2 Cell Lines-Using Lipofectin (Life Technologies, Inc.) Drosophila S2 cells were transfected with pRmHa3 constructs. The pRmHa3 constructs expressing the neuronal form of human L1-CAM and human TAG-1 protein, respectively, have been described previously (11,23). Subcloned cDNAs under the control of the Drosophila metallothionein promoter (24) can be expressed by the addition of 0.7 mM CuSO 4 to the cell culture medium. Co-transfection with the pPC4 plasmid was performed to confer ␣-amanitin resistance as a selectable marker to transfected cells (25). Detailed methods for establishing cloned S2 cell lines using soft agar cloning have been previously reported in detail (26). Individual cell clones were induced overnight and analyzed by Western blotting for high expression of the transfected cDNAs. Selected lines, designated S2:L1-CAM, S2:TAG-1, and S2:L1/ TAG-1, expressed either the neuronal form of human L1-CAM, human TAG-1, or both proteins (Fig. 1).
Cell Aggregation Assays-To discriminate between homophilic and heterophilic cell adhesion mechanisms, mixing experiments were performed using two transfected cell lines, one of which was labeled with the vital fluorescent membrane dye DiI (Molecular Probes, Eugene, OR) as described by Hortsch et al. (27). Labeled cells were mixed with unlabeled cells at a ratio of 1:1 to a final cell concentration of 3 ϫ 10 6 cells/ml. Mixed cell populations were induced overnight by the addition of 0.7 mM CuSO 4 and aggregated for 4 h at room temperature on a shaking platform at 200 rpm. Cell aggregates containing at least 10 cells were analyzed by phase contrast and epifluorescence microscopy on a Leitz Fluovert microscope for the presence or absence of fluorescently labeled cells. Evaluating more than 100 cell aggregates per experiment, the percentage of cell clusters with five or more DiI-labeled cells was calculated for each combination of labeled and unlabeled cells.
SDS-PAGE and Western Blot Analysis-Transfected S2 cells were pelleted and solubilized in SDS-containing buffer. Total cell proteins were separated by electrophoresis in 10% SDS-polyacrylamide gels and transferred onto nitrocellulose filters. Subsequently, the blots were probed with specific primary and horseradish peroxidase-conjugated secondary antibodies and developed with 3,3Ј-diaminobenzidine as described by Hortsch et al. (28).
Immunofluorescence Staining and Confocal Microscopy-Immunocytochemistry of Drosophila ankyrin distribution in S2 cells expressing different molecules was performed as described previously (10). Briefly, cells expressing TAG-1 were mixed with cells expressing both TAG-1 and L1-CAM at a ratio of 20:1 and incubated overnight with 0.7 mM CuSO 4 . Subsequently, the cells were allowed to aggregate for 30 min on a shaking platform before being attached to polylysine-coated microscope slides. Attached cells were fixed with 2% paraformaldehyde for 10 min and permeabilized for 10 min with 0.5% Triton X-100 in TBS buffer (10 mM Tris/HCl, pH 7.5, 0.15 M NaCl, and 5% newborn calf serum). Rabbit anti-L1-CAM and mouse anti-Drosophila ankyrin were used as primary antibodies, followed by an incubation with fluorescein isothiocyanate-or Texas Red-labeled secondary antibodies. Slides were viewed with a Bio-Rad MRC 600 confocal scanning laser microscope.
Immunoprecipitation of L1-CAM and TAG-1-Immunoprecipitations were performed using a modification of the protocol by Anderson and Blobel (29). For each immunoprecipitation, 1 ϫ 10 7 cells expressing either L1-CAM or both L1-CAM and TAG-1 were induced overnight with 0.7 mM CuSO 4 . Without inducing cell aggregation by an incubation step on a rotary shaker the cells were pelleted and solubilized in cold dilution buffer (60 mM Tris/HCl, pH 7.5, 80 mM NaCl, 1.25% Triton X-100, 6 mM EDTA, and a mixture of protease inhibitors). The soluble fraction was incubated overnight with rabbit anti-TAG-1 antibodies. Supernatants were further incubated with Protein A-Sepharose beads (Amersham Pharmacia Biotech) for 2 h, and immunoprecipitates were eluted with SDS gel electrophoresis buffer after three washing steps. After separation on 10% SDS-PAGE gels, proteins were transferred to a nitrocellulose filter and then probed with anti-L1-CAM antibody.

Drosophila S2 Cells Which Express Human TAG-1 or L1-CAM Do Not Form Mixed
Aggregates-Human L1-CAM and TAG-1 have both been shown to induce the aggregation Drosophila S2 cells by a homophilic mechanism (23,27). Although initial experiments using the chicken homologues of TAG-1 and L1-CAM, axonin-1 and Ng-CAM, coated to Covaspheres, suggested that these molecules might also be able to interact with each other when expressed on different surfaces (30), follow-up experiments using axonin-1 and Ng-CAM expressed by tissue culture cells indicated that these molecules are unable to engage in an adhesive trans-interaction (18).
To investigate whether human L1-CAM and TAG-1 are able to induce the aggregation of cells by a heterophilic mechanism, populations of S2 cells expressing either L1-CAM, TAG-1, or the unrelated CAM Drosophila fasciclin I were stained with the fluorescent dye DiI and mixed with unlabeled S2 cells expressing the same set of homophilic adhesion molecules. Most cell aggregates were comprised of several hundred cells, but no significant inclusion of cells expressing TAG-1 into cell clusters expressing L1-CAM and vice versa was observed (Fig. 2). As a result approximately 50% of all cell clusters consisted of DiIlabeled cells expressing one adhesive molecule, whereas the other clusters consisted of unlabeled cells expressing the other CAM. Similar results were obtained mixing induced S2:L1-CAM or S2:TAG-1 cells with S2 cells expressing the unrelated CAM, fasciclin I. Drosophila fasciclin I was used as a negative control in these experiments, because it is not known to interact with L1-type molecules and a TAG-1 homologue has not been identified in Drosophila. In contrast, more than 90% of all cell clusters contained labeled as well as unlabeled S2 cells when both populations expressed either human L1-CAM or TAG-1. This is in agreement with the homophilic adhesive properties that have been demonstrated for both molecules (23,27). These results also indicate that cells expressing either human TAG-1 or L1-CAM do not form mixed aggregates and that these two adhesive molecules do not engage in a measurable heterophilic trans-interaction with each other.
Human L1-CAM and TAG-1 Engage in a Cis-interaction within the Plane of the Plasma Membrane-Stable S2 cell lines expressing both L1-CAM and TAG-1 were used to demonstrate that the two molecules interact when expressed in the same cell. Immunoprecipitations with anti-TAG-1 antibodies followed by immunoblotting with an anti-L1-CAM antiserum indicated that L1-CAM and TAG-1 form stable complexes when present in the same plasma membrane (Fig. 3, lane 2). As shown in lane 1, no L1-CAM was immunoprecipitated from these cells with non-immune rabbit serum. The observed coimmunoprecipitation of L1-CAM with TAG-1 was also not caused by a cross-reactivity of the anti-TAG-1 antiserum toward human L1-CAM, because this antiserum failed to precipitate L1-CAM from cells that did not express TAG-1 (Fig. 3,  lane 3).

FIG. 3. Co-immunoprecipitation of L1-CAM and TAG-1 protein from S2 cell protein extracts. S2 cells expressing either human
L1-CAM or both human L1-CAM and human TAG-1 were induced overnight with 0.7 mM CuSO 4 and solubilized with a Triton X-100containing buffer. As indicated at the bottom of the figure immunoprecipitations from these extracts were performed using either a rabbit non-immune serum or a rabbit antiserum against TAG-1. Immunoprecipitated proteins were separated on 10% SDS-PAGE gels and transferred to nitrocellulose filters, which were probed with a rabbit anti-L1-CAM antiserum.

FIG. 4. Ankyrin is recruited to cell contact sites in S2 cells expressing human L1-CAM but not in cell aggregates expressing human TAG-1.
After protein induction from the transfected cDNA constructs S2 cells expressing human L1-CAM (A and B) or human TAG-1 (C and D) were allowed to aggregate, fixed, and fluorescently stained using a mouse anti-Drosophila ankyrin antiserum. Scale bar is 25 m.

FIG. 5. Ankyrin recruitment to human L1-CAM is induced by homophilic TAG-1-mediated cell adhesion.
S2 cells expressing human TAG-1 protein were mixed at a ratio of 20:1 with S2 cells coexpressing human TAG-1 and human L1-CAM (A-D) or with S2 cells expressing only human L1-CAM (E and F). This ratio of cells was selected to maximize the number of cell contacts between cells expressing both adhesion molecules with cells expressing only TAG-1, rather than contacts between cells expressing L1-CAM and TAG-1. After protein induction cells were briefly allowed to aggregate and subsequently processed for double immunofluorescence using a rabbit anti-L1-CAM antiserum (left panels) and a mouse anti-Drosophila ankyrin antiserum (right panels). The arrow in A marks a cell contact between two cells expressing L1-CAM as well as TAG-1. Cell contacts between cells expressing TAG-1 with cells expressing both adhesion molecules are indicated by arrowheads. Scale bar is 25 m.
Homophilic, TAG-1-mediated S2 Cell Aggregation Does Not Induce Ankyrin Recruitment at Cell Contact Sites-Members of the L1 family induce the specific recruitment of ankyrin and other components of the membrane skeleton to sites of cell contact in S2 cell aggregates (10,11). As shown for human L1-CAM in Fig. 4, A and B, endogenous S2 cell ankyrin is specifically recruited to cell contact sites by L1-CAM-mediated cell adhesion. No ankyrin staining of the plasma membrane can be detected in non-contact areas or in cells that have not joined cell aggregates. S2 cells expressing human TAG-1 exhibit robust homophilic cell aggregation (Fig. 2). However, no ankyrin recruitment to cell contacts or to other areas of the plasma membrane was ever observed in TAG-1-expressing cell clusters (Fig. 4, C and D). In some cells, especially cells exhibiting no ankyrin recruitment to cell contact sites, ankyrin staining appears in a punctate pattern. The reason for this punctate intracellular distribution of ankyrin in S2 cells is unknown but has been observed and described before (10,12).

TAG-1 Homophilic Cell Adhesion Induces L1-mediated
Ankyrin Recruitment-Because TAG-1 is unable to directly interact with the intracellular membrane skeleton, we tested the possibility that it might activate the ability of L1-CAM to bind ankyrin. No recruitment of ankyrin to the plasma membrane was observed in single, non-aggregated S2 cells expressing L1-CAM as well as TAG-1 (not shown). However, when these co-expressing cells were mixed and co-aggregated with S2 cells expressing only TAG-1, a strong recruitment of ankyrin was observed at cell contact sites between these two cell lines (Fig. 5, A-D). The example shown in A depicts two cells that stain positive for L1-CAM (left side) and therefore express both L1-CAM and TAG-1. The corresponding micrograph on the right side of A shows the distribution of ankyrin in these cells.
As indicated by the arrow ankyrin was recruited to the cell contact between the two L1-CAM-positive cells. Ankyrin was also recruited to contact sites these two cells had developed to cells expressing only TAG-1 (marked by arrowheads). B, C, and D show a range of other examples of ankyrin recruitment to cell contact sites between the two different S2 cell types. A quantitative evaluation of these experiments indicates that approximately 75% of such cell contacts exhibited a recruitment of ankyrin (Table I).
Although we observed no trans-interaction between TAG-1 and L1-CAM in our S2 cell aggregation experiment shown in Fig. 2, we considered that a weak trans-interaction between TAG-1 and L1-CAM expressed by two different cells might be responsible for the observed induction of ankyrin binding to L1-CAM. We therefore determined the ankyrin distribution in mixtures of cells expressing either TAG-1 or L1-CAM. By analyzing a large number of immunostained slides, we identified FIG. 6. Model of homo-and heterophilic interactions of TAG-1. The homophilic trans-interaction between TAG molecules expressed by different cells is mediated by its fibronectin type III domains (23), whereas the Ig domains of TAG appear to be responsible for its heterophilic cis-interaction to L1-type proteins (32). The protein domains of L1-CAM that are involved in binding to TAG-1 are currently unknown. The two TAG-1 molecules are drawn in a horseshoe-shaped domain arrangement, which has been proposed for the chicken homologue of TAG-1, axonin-1 (32).

TABLE I Quantification of ankyrin recruitment to cell contact sites involving cells expressing TAG-1 and cells expressing either human L1-CAM or human L1-CAM and TAG-1
Type of cell contact rare instances in which a TAG-1-expressing cell ended up in contact with a cell expressing just L1-CAM (Fig. 5, E and F). In none of the 29 examples we analyzed did we observe a recruitment of ankyrin to these cell contact points (Fig. 5 and Table I). This indicates that L1-CAM and TAG-1 must be co-expressed in the same cell for TAG-1-mediated cell adhesion to induce ankyrin binding to L1-CAM.

DISCUSSION
Complex homo-and heterophilic interactions between different adhesion molecules within the plane of the plasma membrane and between adjacent cells have been proposed to play an important role in regulating the growth and guidance of axons during embryonic neurogenesis (5). Our present in vitro study addresses how two different CAMs, human TAG-1 and human L1-CAM, interact and functionally regulate each other. Some groups of CAMs are transmembrane proteins and therefore potentially able to directly influence intracellular processes, e.g. by activating second messenger signaling cascades and/or by reorganizing components of the cytoskeleton. However, other CAM families, such as the TAG-1/axonin-1 and the F3/ F11 groups, are anchored in the plasma membrane by a GPI moiety and therefore lack the means to interact with cytoplasmic proteins without additional linker proteins. Nevertheless, GPI-anchored CAMs are also able to operate as signal-transducing molecules during neuronal development. Although TAG-1/axonin-1 is fully functional as a homophilic CAM without engaging in any heterophilic cis-interactions, its ability to associate with other membrane proteins, especially L1-CAM, appears to be essential for its neurite outgrowth-promoting function (17,31). Several models similar to the one displayed in Fig. 6 have been proposed in which TAG-1/axonin-1 associates with other membrane proteins expressed in the same plasma membrane in a cis-interaction to form a signal-transducing, multimeric protein complex (17,18). A similar model has also been suggested to explain the adhesion-dependent activation of neuronal FGF receptors by L1-CAM, NCAM, and N-cadherin (8).
Although several molecular details of the TAG-1 interaction with L1-CAM have been elucidated, its regulatory and functional aspects are not well understood. L1-CAM/Ng-CAM not only can be co-immunoprecipitated with TAG-1/axonin-1, but cross-linking experiments suggest that both molecules directly bind to each other (18,20). The part of the axonin-1 molecule responsible for this cis-interaction has been mapped to its amino-terminal Ig protein domains (32), whereas the homophilic adhesive activity of TAG-1/axonin-1 is mediated by its fibronectin type III domains (23). The protein domains of L1-CAM that mediate TAG-1 recognition and binding have not been identified yet. Also whether the homophilic adhesive activities of L1-CAM and TAG-1 regulate their cis-interaction is currently unknown. However, co-capping experiments reported by Buchstaller et al. (18) and the quantity of L1-CAM that co-immunoprecipitated with TAG-1 in our experiments suggest that a significant fraction of L1-CAM molecules forms heterodimeric complexes with TAG-1 before the two CAMs engage in cell adhesion and L1-CAM binds to the membrane skeleton. After L1-CAM interacts with ankyrin, it becomes resistant to Triton X-100 extraction as used in our immunoprecipitation experiments and is unavailable for precipitation with antibodies (10).
TAG-1/axonin-1 and L1-CAM/Ng-CAM are co-expressed in several locations during nervous system development, suggesting that their interaction is physiologically relevant (31,33,34). In cultures of chicken dorsal root ganglion neurons Ng-CAM and axonin-1 protein are found in overlapping areas on growth cone membranes, and axonin-1 expression by these cells is required for neurite outgrowth on both axonin-1 and Ng-CAM substrata (31). In contrast, in other developing neurons, e.g. in the rat embryonic spinal cord, TAG-1 and L1-CAM expression appear to be locally and temporally segregated (22,35). This finding indicates that TAG-1/axonin-1 and L1-CAM/ Ng-CAM are not obligatory co-receptors but are able to function independently. It might mean either that the signaling capabilities of TAG-1/axonin-1 are limited in certain areas of the developing nervous system or at certain developmental time points or that TAG-1/axonin-1 associates with other membrane proteins than L1-CAM/Ng-CAM to form functionally active protein complexes.
Although the cis-activation of neuronal FGF receptors by L1-CAM is thought to be the initial step in adhesion-induced neurite outgrowth, processes such as the recruitment of other signaling molecules to cell contact sites might also have an important, more indirect role. Our findings presented here would support the hypothesis that TAG-1 homophilic adhesion might also activate neuronal FGF receptor activity via L1-CAM. Other cellular changes that appear to be regulated by the TAG-1/axonin-1 interaction with L1-CAM/Ng-CAM are the recruitment and the release of several different protein kinase activities that are associated with the two CAMs in chicken dorsal root ganglion neurons (20). In the case of integrin-and cadherin-mediated cell adhesion, interactions with cytoskeletal elements induce the assembly of signal-generating and -processing multiprotein complexes at the cytoplasmic aspect of the adhesion contact site. This makes the observed assembly of membrane skeleton components in response to TAG-1-mediated homophilic cell adhesion especially significant. However, the most tantalizing aspect of the results reported here is the observation that TAG-1 homophilic cell adhesion directly regulates the functional status of L1-CAM. These findings suggest a mechanism that explains how GPI-anchored CAMs might actively participate in regulating the growth, organization, and differentiation of neuronal cells during development.