Receptor Clustering Drives Polarized Assembly of Ankyrin

Expression of the L1 family cell adhesion molecule neuroglian in Drosophila S2 cells leads to cell aggregation and polarized ankyrin accumulation at sites of cell-cell contact. Thus neuroglian adhesion generates a spatial cue for polarized assembly of ankyrin and the spectrin cytoskeleton. Here we characterized a chimera of the extracellular and transmembrane domains of rat CD2 fused to the cytoplasmic domain of neuroglian. The chimera was used to test the hypothesis that clustering of neuroglian at sites of adhesion generates the signal that activates ankyrin binding. Abundant expression of the chimera at the plasma membrane was not a sufficient cue to drive ankyrin assembly, since ankyrin remained diffusely distributed throughout the cytoplasm of CD2-neuroglian-expressing cells. However, ankyrin became highly enriched at sites of antibody-induced capping of CD2-neuroglian. Spectrin codistributed with ankyrin at capped sites. A green fluorescent protein-tagged ankyrin was used to monitor ankyrin distribution in living cells. Enhanced green fluorescent protein-ankyrin behaved identically to antibody-stained endogenous ankyrin, proving that the polarized accumulation of ankyrin was not an artifact of fixing and staining cells. We propose a model in which clustering of neuroglian induces a conformational change in the cytoplasmic domain that drives polarized assembly of the spectrin cytoskeleton.


Introduction
transgenes and processed for live cell capping with primary and secondary antibodies two days later, as described above. Live cells were viewed by epifluorescence, photographed using an ausJena Jenalumar photomicroscope. Images were obtained on photographic film, digitized using a Polaroid SprintScan scanner and edited using Adobe Photoshop 5.0.
Western blotting was carried out as previously described (24). Total S2 cell proteins (

Quantification of capping experiments
Capped S2 cells expressing CD2/nrg or CD2 were scored for the expression of fluorescent antibody stained caps using the rhodamine channel. To score primary antibody-treated cells, cells were fixed prior to addition of fluorescent secondary antibody. Fifty cells were scored for ankyrin or spectrin colocalization at caps in each experiment and results were expressed as a percentage of total capped cells.
Control uncapped cells were scored for the presence of conspicuous CD2 staining, then scored for ankyrin colocalization.

Results
Expression of the adhesion molecule neuroglian in Drosophila S2 tissue culture cells leads to the formation of large aggregates of cells as well as the recruitment of ankyrin specifically to sites of cell-cell contact (13). To study the ankyrin recruitment process independently of cell-cell adhesion, we engineered a chimeric protein composed of the extracellular and transmembrane domains of rat CD2 fused to the cytoplasmic sequence of the nrg 167 isoform of neuroglian (Fig. 1) Expression of recombinant CD2 molecules and their effects on assembly of ankyrin in S2 cells were examined by immunofluorescent staining of transiently transfected populations. Typically less than 50% of the cells in the population expressed recombinant CD2, and among these there was a broad range of expression level (data not shown). Cells were fixed and permeabilized for antibody staining.
Conspicuous uniform plasma membrane staining and variable intracellular staining were observed in cells expressing either CD2/nrg or CD2 (Fig. 2, column 1). There was no detectable effect of CD2 expression on the intracellular distribution of ankyrin (column 2). Ankyrin staining was found in a variable speckled pattern throughout the cytoplasm of S2 cells, whether or not the cells were transfected. Thus the presence of the neuroglian cytoplasmic domain in S2 cells was not sufficient to bring about a detectable redistribution of ankyrin to the plasma membrane.
The distribution of ankyrin was also examined after capping CD2 by addition of anti-CD2 antibody to living cells. Capping was induced by the sequential addition of monoclonal anti-CD2 antibody and fluorescent goat anti-mouse secondary antibody. Addition of antibody to living cells resulted in extensive capping of both CD2/nrg and CD2 (Fig. 2, column 3). Cap morphology varied from numerous small islands of staining (arrows) to large polar caps (arrowhead). After capping, cells were fixed and permeabilized for staining with rabbit anti-ankyrin antibody (Fig. 2, column 4). CD2/nrg expressing cells exhibited a dramatic redistribution of ankyrin to capped sites (Fig. 2, E-F), although not all caps were associated with detectable ankyrin staining (e.g. The effects of CD2/nrg expression on ankyrin distribution were quantified in large populations of transfected cells (Fig. 3). The fraction of cells with caps that exhibited colocalization of ankyrin was expressed as a percentage of total capped cells (n = 50). Forty percent of cells expressing the CD2/nrg chimera exhibited prominent ankyrin colocalization. About 20% of cells treated with primary antibody alone also exhibited recruitment of ankyrin to cap sites. Treatment of cells with monoclonal primary antibody alone was expected to induce formation of CD2 dimers rather than caps. However, antibody binding appears to induce a conformational change in CD2 that favors lateral clustering of molecules in the plane of the membrane (26). Thus, while caps formed by primary antibody alone were not as prominent or compact as those induced with primary and secondary antibodies (data not shown), there was a conspicuous redistribution of CD2/nrg as well as ankyrin in primary antibody-treated cells that was distinct from non-antibody-treated control cells.
CD2/nrg capping was also expected to induce a redistribution of spectrin, since ankyrin is the adapter protein that links spectrin to the plasma membrane. There was a dramatic redistribution of spectrin in capped CD2/nrg cells (Fig. 4A-B). The spectrin in control cells (not shown) and capped CD2 cells (Fig. 4C-D) was found in a diffuse distribution in the cytoplasm and in numerous perinuclear puncta, as previously described (27). Quantification of the spectrin redistribution revealed that it was not as efficient as ankyrin redistribution (Fig. 3, right), but it was equally dependent on the cytoplasmic domain of neuroglian and capping. Thus antibody capping of CD2/nrg induced assembly of a membrane cytoskeleton complex of ankyrin and spectrin comparable to what was previously observed in S2 cells expressing authentic neuroglian (13).
The response of ankyrin to neuroglian expression was also monitored in cells expressing a recombinant ankyrin-green fluorescent protein (EGFP) construct (Fig. 5). It was important to establish that the patterns of ankyrin staining in S2 cells were not an artefact of fixing and/or permeabilizing the cells. The NcoI start site of EGFP was fused to a modified ankyrin coding sequence in which the stop codon was replaced with an NcoI site. The ankyrin template used also included an amino-terminal myc epitope tag (13). Expression of the desired product was confirmed in Western blots of transfected cells . Thus the ankyrin-EGFP product was abundantly expressed and stable in S2 cells.
The ankyrin-EGFP marker was monitored in living, neuroglian-expressing S2 cells by confocal microscopy. Ankyrin-EGFP was excluded from the nucleus but was otherwise diffusely distributed throughout the cytoplasm of non-aggregated cells (not shown). A confocal through-focus series revealed that ankyrin-EGFP was dramatically redistributed to sites of cell-cell contact in neuroglian expressing cells (Fig. 6A, arrow). No ankyrin-EGFP above the background of cytoplasmic staining was detected at non-contact regions of the plasma membrane (arrowhead). Thus the distribution of ankyrin-EGFP in living cells precisely matched the previously described distribution of ankyrin in antibody-stained cells (13). living S2 cells (Fig. 6B). Cells were treated with primary and secondary antibody, as described above, then directly viewed and photographed while still alive. Antibody treatment caused a dramatic redistribution of CD2/nrg and CD2 into caps (Fig. 6B, arrows). Ankyrin-EGFP codistributed with caps in cells expressing the chimera, but not in cells expressing truncated CD2 (Fig. 6C, arrows). Ankyrin-EGFP was not detectably associated with the plasma membrane prior to antibody-induced capping (data not shown). Thus the dramatic enrichment of ankyrin at sites of antibody-induced capping of CD2/nrg and at sites of neuroglian-mediated cell-cell adhesion was not an artefact of the processing steps used in antibody labeling experiments.

Discussion
Previous studies established that the cytoplasmic domain of neuroglian recruits ankyrin to sites of cell-cell adhesion in neuroglian-expressing S2 cells (13,15). Several possible explanations for the polarizing effect of neuroglian on ankyrin were proposed, including 1) mobilization of neuroglian to cell contacts, thereby sequestering the ankyrin binding site; 2) transmission of an allosteric signal that selectively activates ankyrin binding by neuroglian molecules engaged in extracellular adhesion; and 3) inhibition of ankyrin binding to neuroglian at non-contact sites by tyrosine phosphorylation within the ankyrin-binding site of neuroglian. However, each of these mechanisms was ruled out in previous studies of wild type and mutant neuroglian molecules (13,15). A fourth mechanism, local activation of ankyrin recruitment by clustering of neuroglian molecules in the plane of the membrane, had not been tested. The finding here that ankyrin codistributes with a CD2-neuroglian chimera only after antibody-induced capping provides compelling evidence that ankyrin binding to neuroglian is modulated by receptor clustering.
It was important to establish that the apparent concentration of ankyrin at sites of neuroglian clustering was not an artefact of either cell aggregation or the processing steps used in antibody staining experiments. The CD2/nrg chimera used here made it possible to uncouple ankyrin recruitment from cell adhesion and to exert precise control over the timecourse of ankyrin recruitment. The chimera could be expressed at high levels without recruiting ankyrin to the plasma membrane. It was not possible to express authentic neuroglian without inducing aggregation and ankyrin recruitment once neuroglian accumulated to a sufficient level. The fixation and permeabilization steps before antibody staining were a source of further concern. Most of the neuroglian in cell aggregates was present at non-contact regions of the plasma membrane, but was extracted during permeabilization with detergent (13). Consequently, ankyrin and neuroglian both appeared to be concentrated at cell contacts in antibody-stained cells. Use of the ankyrin-EGFP fusion described here made it possible to monitor the distribution of ankyrin without cell processing artefacts. Ankyrin-EGFP strictly codistributed with sites of neuroglian-mediated adhesion and not with the distribution of total neuroglian.
The behavior of ankyrin-EGFP in CD2/nrg-expressing S2 cells was remarkably different from the behavior of EGFP-tagged ankyrin in mammalian cells. When ankyrin G -EGFP was expressed in human kidney 293 cells it was diffusely distributed throughout the cytoplasm and was not detectably associated with the plasma membrane. When co-expressed with intact neurofascin, ankyrin G -EGFP was efficiently recruited to the plasma membrane, whether or not the cells formed aggregates (28). Thus the recruitment of ankyrin to the plasma membrane by rat neurofascin was independent of cell adhesion. This is in marked contrast to Drosophila ankyrin-EGFP which only associated with the plasma membrane at cell contacts in neuroglian-expressing cells or with antibody-induced caps in CD2/nrg-expressing cells (this study). The difference in the ankyrin recruiting activities of the two proteins was unexpected given the sequence conservation of their ankyrin binding sites (70% identity over 27 residues; Fig. 6). One question concerns the valency of the ankyrin interaction with L1 family members. Molecular and biochemical studies indicate that there are two binding sites for neurofascin in mammalian ankyrins and that a single ankyrin can simultaneously interact with two neurofascin molecules, as well as band 3 (29). However, there is no direct evidence that both sites are utilized or required in situ. The model posits that a single ankyrin molecule associates with the cytoplasmic domains of two neuroglian or neurofascin molecules, although 1:1 binding of ankyrin to neuroglian would also serve to mask charges. We attempted to address the valency of ankyrin in the present study without success. CD2/nrg was coexpressed with authentic neuroglian in S2 cells to ask if recruitment of ankyrin to cell contacts by neuroglian would lead to a simultaneous recruitment of CD2/nrg. CD2 staining was observed at cell contacts (data not shown), but the same result was obtained with CD2, indicating that recruitment to cell contacts occurred via the extracellular domain of CD2, rather than the ankyrin-binding site of neuroglian.
Thus other experimental approaches, perhaps using different neuroglian chimeras, will be required to determine the number of neuroglian molecules that associate with ankyrin in situ.
Neuroglian molecules are likely to be clustered at cell contact sites by diffusion-mediated trapping. Clustering of adhesion molecules at adhesive sites by this mechanism is thought to enhance the strength and stability of cell contacts (30-32). The model described here suggests that clustering of neuroglian is also responsible for the selective recruitment of ankyrin to cell contacts. In addition, the model provides a rationale to explain inside-out regulation of neuroglian mediated adhesion (15,33). The model proposed here suggests that ankyrin binding to neuroglian is activated by receptor clustering. An alternative interpretation of the data is that ankyrin is constitutively bound to neuroglian in the absence of clustering, but the concentration of ankyrin at unclustered sites is below its threshold of detection. Clustering of neuroglian could conceivably elevate the ankyrin concentration above that threshold. This alternate explanation seems unlikely for a number of reasons. First, threshold of detection was not an issue with mammalian ankyrin-EGFP which was dramatically redistributed to the plasma membrane in response to neurofascin expression (28). Drosophila ankyrin labeled with the same EGFP tag did not codistribute with neuroglian at non-contact regions of the plasma membrane in neuroglianexpressing cells or with uncapped CD2/nrg. Second, not all of the caps formed by antibody treatment of CD2/nrg expressing cells were positive for ankyrin. This result is more consistent with an induced association that is moderately efficient as opposed to accretion of ankyrin that is constitutively associated with neuroglian. In the latter case, ankyrin staining should always be observed when its receptor is clustered. Third, ankyrin did not colocalize with neuroglian at non-contact regions of the S2 cell plasma membrane under conditions where detection of ankyrin relative to neuroglian was not limiting (13).
Antibody staining experiments suggest that the enrichment of neuroglian staining at cell contacts is relatively small compared to the dramatic enrichment of ankyrin at contacts. We suggest that the enrichment of ankyrin is caused by an increase in the affinity of the ankyrin-neuroglian interaction that is brought about by receptor clustering.
There are likely to be additional activating mechanisms that operate during assembly of the cytoskeleton, even within the L1 family of proteins. In the nervous system, spectrin, ankyrin, sodium channels and neurofascin all codistribute at the node of Ranvier (34). However, targeting to the node does not require cell-cell adhesion since these molecules also form node-like clusters in cultured neurons in response to a protein factor secreted by oligodendrocytes (35). In light of this result, it is not surprising that the recruitment of ankyrin to the plasma membrane by neurofascin in 293 cells was independent of cell-cell adhesion. One possibility is that recruitment of ankyrin by neurofascin at the node is regulated by tyrosine phosphorylation, which is known to have a potent modulatory effect on ankyrin binding in vitro