NrCAM Coupling to the Cytoskeleton Depends on Multiple Protein Domains and Partitioning into Lipid Rafts
NrCAM is a cell adhesion molecule of the L1 family that is implicated in the control of axonal growth. Adhesive contacts may promote advance of the growth cone by triggering the coupling of membrane receptors with the F-actin retrograde flow. We sought to understand the mechanisms leading to clutching the F-actin at the site of ligand-mediated clustering of NrCAM. Using optical tweezers and single particle tracking of beads coated with the ligand TAG-1, we analyzed the mobility of NrCAM-deletion mutants transfected in a neuroblastoma cell line. Deletion of the cytoplasmic tail did not prevent the coupling of NrCAM to the actin flow. An additional deletion of the FNIII domains to remove cis-interactions, was necessary to abolish the rearward movement of TAG-1 beads, which instead switched to a stationary behavior. Next, we showed that the actin-dependent retrograde movement of NrCAM required partitioning into lipid rafts as indicated by cholesterol depletion experiments using methyl-β-cyclodextrin. Recruitment of the raft component caveolin-1 was induced at the adhesive contact between the cell surface and TAG-1 beads, indicating that enlarged rafts were generated. Photobleaching experiments showed that the lateral mobility of NrCAM increased with raft dispersion in these contact areas, further suggesting that TAG-1–coated beads induced the coalescence of lipid rafts. In conclusion, we propose that anchoring of NrCAM with the retrograde actin flow can be triggered by adhesive contacts via cooperative processes including interactions with the cytoplasmic tail, formation of cis-complex via the FNIII repeats, and lipid raft aggregation.
Cell adhesion molecules of the immunoglobulin superfamily (IgCAMs) are major players in the control of axonal growth and guidance. NrCAM is a transmembrane molecule that contains both Ig and fibronectin type III (FNIII) domains and belongs to the L1-family of IgCAMs also including L1, CHL1, and neurofascin. In vitro studies have indicated that NrCAM participates to diverse combinatorial associations of IgCAMs and extracellular matrix molecules that finely tune axonal growth. NrCAM mediates homophilic binding and heterophilic interactions with F3/contactin, TAG-1, neurofascin, RPTPβ/z (Volkmer et al., 1996; Sakurai et al., 1997; Zacharias et al., 1999). NrCAM has been involved in the axogenesis of different neuronal populations acting both as a ligand and a receptor. It is implicated in the control of sensory axon patterning and in the midline crossing of commissural axons in the spinal cord of chick embryo (Stoeckli and Landmesser, 1995; Perrin et al., 2001). NrCAM acts as a functional receptor for F3/contactin to regulate axonal outgrowth of cerebellar neurons (Faivre-Sarrailh et al., 1999; Sakurai et al., 2001; Falk et al., 2002). NrCAM-deficient mice do not display any gross morphological alteration, but double knock-out mice for L1 and NrCAM exhibit severe defects in cerebellar morphogenesis, indicating that the two molecules may have overlapping functions (Sakurai et al., 2001).
The growth cone is the sensor and motor device on which the trajectory of the growing axon relies. Growth cone motility depends on the dynamics of actin filaments, generated by actin polymerization at the leading edge and myosindependent rearward flow of F-actin. Extension or retraction of the growth cone results from a balance between these two opposing processes that appear to be independently regulated by environmental cues (Lin and Forscher, 1995; Jay, 2000). Substrate-bound cues may promote advance of the growth cone by physically anchoring membrane receptors with the actin retrograde flow, thereby generating tension forces needed to pull the growth cone forward. This was first evidenced in neurons for the Aplysia homologue of NCAM, ApCAM (Suter et al., 1998). Coupling of IgCAMs with the actin flow has been also established for NrCAM (Faivre-Sarrailh et al., 1999) and L1 (Kamiguchi and Yoshihara, 2001). Such linkage between IgCAMs and the actin filaments may consist in a general transductory mechanism for axonal guidance and has been also evidenced for other classes of membrane receptors, the β1-integrin and N-cadherin (Schmidt et al., 1995; Grabham et al., 2000; Lambert et al., 2002). The question arises of whether distinct types of adaptors mediate the recruitment of F-actin by the different families of membrane receptors.
Several motifs of interactions with the cytoskeleton have been identified that are conserved in L1-type molecules. The ankyrin-binding motif has been extensively characterized (Bennett and Chen, 2001) and association with ankyrin-G is essential for the targeting of NrCAM and neurofascin at the node of Ranvier. On the contrary, ankyrin binding does not seem to be a key element for the coupling of L1 with the retrograde actin flow in growth cones. Ankyrin-B440 is only distributed in the proximal region of the growth cone (Kunimoto, 1995; Nishimura et al., 2003), and ankyrin binding has been shown to inhibit retrograde movement of L1 and to mediate a stationary behavior (Gil et al., 2003). The juxtamembrane domain of L1-molecules has been shown to confer association of L1 with the actin cytoskeleton (Dahlin-Huppe et al., 1997). NrCAM also contains a C-terminal PDZ-domain–binding motif. Thus, different domains of the NrCAM cytoplasmic tail can account for the coupling with the actin flow. In addition, cis-interactions between L1-molecules and other transmembrane receptors have been reported (Brümmendorf and Rathjen, 1996). For example, L1-CAM clustering on the cell surface has been shown to induce its homomultimerization and recruitment of integrins via interaction with the FNIII domains (Silletti et al., 2000).
By tracking the movement of microspheres coated with F3/contactin, we previously demonstrated that NrCAM displays an actin-dependent retrograde mobility on the growth cones of cerebellar granule cells (Faivre-Sarrailh et al., 1999). Here, our goal was to further investigate the mechanisms underlying the membrane mobility of NrCAM and its anchorage to the actin cytoskeleton.
MATERIALS AND METHODS
The rat NrCAM12 cDNA is a kind gift of V. Bennett. The HA epitope was inserted by PCR 5 amino acids (aa) downstream of the signal peptide in NrCAM12 subcloned in pIRES1neo (Clontech, Palo Alto, CA). The amino acids of NrCAM12 were numbered as in Davis et al. (1996). A construct with deletion of the cytoplasmic tail (NrCAMΔcyt) was generated by introducing a STOP codon at position 1042. NrCAM-GFP chimeras were generated using pEGFP-C1 (Clontech). The NrCAMΔfn construct, in which the FNIII domains were replaced by the GFP, was generated by PCR amplification of the signal peptide and Ig domain sequence (aa -30–607) and insertion in NheI-AgeI sites upstream of the GFP coding sequence in pEGFP-C1. Next, the transmembrane and cytoplasmic domains (aa 999-1154) were amplified by PCR and inserted in the XhoI-BamHI sites of the polylinker. Alternatively, the NrCAMΔfnΔcyt construct was obtained by PCR amplification of the transmembrane domain (aa 999-1041) and insertion in XhoI-BamHI. The NrCAMΔfnΔCter was obtained by PCR amplification of the transmembrane domain and part of the cytoplasmic tail (aa 999-1084) and insertion in XhoI-BamHI. The GFP-TM construct was generated by insertion of the signal peptide of NrCAM (aa -30–16) in NheI-AgeI sites upstream of GFP and insertion of the transmembrane domain (aa 999-1041) in XhoI-BamHI. The PCR-amplified products were verified by sequencing (Genome Express, Meylan, France).
A recombinant protein was generated using the pQE30 vector (Qiagen, Courtaboeuf, France), which carried a his-tag fused to Ig2–6 domains of rat NrCAM12. After induction, the his-NrCAM fusion protein was purified under denaturing conditions by affinity chromatography on Ni-agarose beads and used for the immunization of rabbits. Rabbit anti-ezrin was a gift from Dr. P. Mangeat. Rabbit anti-GFP antibody and rat anti-HA (3F10) mAb were purchased from Roche (Meylan, France). Rabbit and mouse anti-human Fc, mouse IgM anti–α-actinin, and mouse IgG anti-talin were from Sigma (Saint-Quentin Fallavier, France) and rabbit anti-caveolin-1 from Calbiochem (La Jolla, CA). Alexa 488–conjugated secondary antibodies were purchased from Molecular Probes (Eugene, OR). Peroxidase- and Texas red–conjugated secondary antibodies were from Jackson ImmunoResearch (West Grove, PA).
B104 cells were grown in DMEM medium with 10% FCS. B104 cells were transfected using Lipofectamine Plus (Invitrogen, Carlsbad, CA) with NrCAM, NrCAMΔcyt, NrCAMΔfn, NrCAMΔfnΔcyt, NrCAMΔfnΔCter, or GFP-TM constructs. After 2 weeks of selection with G418, stable lines expressing the different constructs were selected using immunofluorescence for HA, fluorescence of GFP, or binding of the TAG-1IgFc chimera clustered with anti-Fc antibodies.
Cell Extracts and Western Blot
Transfected cells were lyzed for 30 min on ice with 50 mM Tris, pH 7.5, 1% Triton X-100, containing 10 mM MgCl2 and protease inhibitors and then centrifuged at 4°C for 15 min at 15,000 rpm. Cell lysates were analyzed by 7% SDS-PAGE and Western blotting with rabbit anti-NrCAM immune serum (1:1000) or anti-HA mAb (1:1000) and peroxidase-conjugated secondary antibodies using chemiluminescence detection kit (Roche).
Low-density Lubrol-insoluble complexes were prepared as previously described (Faivre-Sarrailh et al., 2000). B104 cells (107) were lyzed in 2 ml of 25 mM MES, pH 6.5, containing 0.15 M NaCl, 1% Lubrol, and protease inhibitors. The extract adjusted to 40% sucrose was placed at the bottom of a 5–30% linear sucrose gradient and centrifuged at 4°C for 17 h at 39,000 rpm in a SW41 Beckman rotor (Fullerton, CA). Fractions (1 ml) were collected and analyzed by 7% SDS-PAGE and immunoblotting for NrCAM constructs using anti-HA or anti-GFP mAb, and for caveolin-1 and α-actinin.
Fc-chimera and Bead Preparation
NrCAMFc, F3Fc, and TAG-1IgFc chimeras were previously described (Faivre-Sarrailh et al., 1999; Pavlou et al., 2002). The chimeras produced in eukaryotic cells were immunoadsorbed on mouse or rabbit anti-human Fc-coupled latex sulfate beads of 1-μm (Polysciences, Niles, IL) or 4-μm (Frontier Bioscience, Interfacial Dynamics, Portland, OR) diameter as described previously (Faivre-Sarrailh et al., 1999). For concanavalin-A (con-A) coating, carboxylated beads (Polysciences) were covalently coupled to biotinylated BSA, incubated with Neutravidin (Molecular Probes), and then with biotinylated con-A (Sigma) as described (Choquet et al., 1997).
Videomicroscopy, Single-particle Tracking, and Laser Trapping
Cells were plated on poly-l-lysine coated glass coverslips (10 μg/ml for 1 h at 37°C) the day before the experiment. To enhance cell motility, cells were serum-starved for 1 h before being mounted in an observation chamber with DMEM, 1% BSA, and 20 mM HEPES containing diluted TAG-1 beads. When indicated, 2 μM cytochalasin D, 10 mM methyl-β-cyclodextrin (MBCD), or a mixture of cholesterol and MBCD at a molar ratio of 1:6 (Kenworthy et al., 2004) was added to the medium at this point. The optical setup consisted of an inverted microscope (Olympus IX 70, Tokyo, Japan) fed through its epifluorescence port by a Nd:YAG laser beam (Compass 1064 nm, Coherent, Palo Alto, CA), reflected by a dichroic mirror (Chroma, Rockingham, VT) and focused by a 100×/1.45 NA oil objective (Zeiss, Le Pecq, France). The beam power was adjusted at 150 mW (we measured ∼10 μW at the front of the objective, which is small enough not to harm the cells). Microbeads were trapped with the laser beam and manipulated using a motorized stage (MarzHauser, Wetzlar, Germany), and digital images were acquired with DIC illumination using a cooled CCD camera (HQ Cool Snap, Roper Scientific, Evry, France). Beads were placed for 5 s in the peripheral region of cell lamellipodia. Behavior of the beads was observed for 1 min after initial attachment and classified as rearward moving or not. Resistance to the laser trap was also tested on each bead, by placing the trap ∼0.5 μm behind the bead, where the trap force is maximal (Choquet et al., 1997). In some experiments nonrearward moving beads were further characterized as diffusing or stationary (if bead displacements did not exceed 1 μm during the 1-min period of observation). Alternatively, bead mobility was recorded at a 40-ms rate for the 90 s after attachment. Bead displacement was followed using the tracking object application of Metamorph software (Universal Imaging Corp, Downington, PA). The mean squared displacement (MSD) was calculated using a homemade software and fitted by the equation 4Dt + V2t2, where t is the time, D the diffusion coefficient, and V the directed velocity (Choquet et al., 1997).
Fluorescence Recovery after Photobleaching
The fluorescence recovery after photobleaching (FRAP) setup used the same microscope together with a 488-nm beam from an argon laser (Innova 300, Coherent). The laser beam is expanded to fill out the pupil of a 100×/1.4 NA objective and adjusted via a translatable lens to yield a diffraction-limited spot on the focal plane. The laser power was set to 150 mW, corresponding to 80 μW at the front of the objective. Wide-field illumination was achieved via a regular Xenon lamp, deported at 90° and joining the optical path via a 70/30 beamsplitter (Chroma). Both the lamp and laser illuminations were controlled by shutters. A region of interest was brought to the position of the laser spot and the FRAP sequence was started: a few reference images were acquired first, and then the sample was bleached by the laser for 2 s, and fluorescence recovery was recorded for 5 min. Images were acquired with a CCD camera with exposure times of 50–200 ms. The whole sequence was driven by a journal written in the Metamorph software. Four FRAP sequences were run per coverslip, bringing the experiment duration to ∼30 min. Using cells fixed with paraformaldehyde, we measured the diameter of the bleached area (∼4 μm for a 2-s exposure at full laser power) and the photobleaching due to the illumination sequence itself (<5%).
Immunofluorescence Staining and Bead-cell Adhesion Assays
MBCD Treatment and Triton X-100 Extraction. Cells were rinsed with serumfree culture medium for 2 h, treated or not with 10 mM MBCD for 30 min, and then incubated with rabbit anti-GFP antibody at 4°C for 30 min. Thereafter, cells were incubated with 0.1% Triton X-100 at 4°C for 3 min, fixed with formaldehyde 4% in PBS for 10 min, and incubated with Texas red–conjugated anti-rabbit immunoglobulins. Incubation with filipin (50 μg/ml) was carried out to detect cholesterol and epifluorescence was done with an UV filter as described in Marchand et al. (2002).
Bead-cell Adhesion Assays. NrCAM- or NrCAMΔfnΔcyt-expressing B104 cells were treated or not with 10 mM MBCD for 30 min and incubated with TAG-1–conjugated beads (4-μm diameter) for 30 min. Cells fixed with 4% paraformaldehyde were permeabilized with Triton X-100 and processed for immunofluorescence staining with anti-caveolin-1 (1:500), anti-ezrin (1:2000), anti–α-actinin (1:200), or anti-talin (1:200) and Texas red–conjugated antirabbit IgG, anti-mouse IgG or anti-mouse IgM, or staining using Texas red–conjugated phalloidin (Molecular Probes). Alternatively, cells were incubated with TAG-1 beads (4 μm) for 15 min and treated or not with 10 mM MBCD for an additional period of 15 min. Then, a quantitative analysis of GFP-fluorescence clusters of NrCAMΔfnΔcyt recruited under the beads was performed on more then 50 cells under each condition and two perpendicular diameters were measured for each GFP cluster using the Metamorph software.
Specificity of Binding between TAG-1 Beads and NrCAM-transfected Cells
Because primary neuronal cultures endogenously express several L1-type molecules, we turned to a simplified cellular model, a rat neuroblastoma B104 cell line (Garver et al., 1997), which does not express endogenous NrCAM (Figure 1B). We established stable lines of B104 cells transfected with full-length NrCAM or several deleted forms of the molecule: deletion of the cytoplasmic tail (NrCAMΔcyt); deletion of the FNIII repeats, which were replaced by GFP (NrCAMΔfn); and deletions of both the FNIII repeats and cytoplasmic tail (NrCAMΔfnΔcyt) or C-terminal region of the cytoplasmic tail (NrCAMΔfnΔCter; Figure 1A). A construct that contains the signal peptide of NrCAM upstream and the transmembrane region of NrCAM downstream of GFP was generated for control experiments (Figure 1A). All constructs were tagged with an HA epitope at the N-terminus. The selected lines transfected with full-length NrCAM or deleted constructs produced similar levels of recombinant protein as analyzed by Western blotting with anti-HA mAb (Figure 1B). Immunofluorescence staining for the HA epitope indicated that all the deletion mutants of NrCAM were expressed at the cell surface (unpublished data).
Because parental B104 cells displayed binding sites for F3/contactin (Figure 1E), we decided to use TAG-1 as another physiological ligand to trigger the coupling of NrCAM to the cytoskeleton. However, TAG-1 displays a strong homophilic binding activity essentially mediated by its FNIII repeats (Tsiotra et al., 1996) that results in the stable aggregation of TAG-1Fc chimera. To favor heterophilic binding, we used a construct deleted from the FNIII repeats (TAG-1IgFc) that has been previously characterized for its strong interaction with NrCAM (Pavlou et al., 2002). Beads coated with the chimera TAG-1IgFc (hereafter abbreviated TAG-1), selectively bound to NrCAM-expressing cells, whereas no significant binding was observed on parental B104 cells (Figure 1, C–E). In addition, the binding sites for soluble TAG-1 were colocalized with the receptor NrCAM expressed at the membrane as illustrated with the NrCAMΔfnΔcyt construct containing the GFP sequence (Figure 1, F–H).
Actin-dependent Retrograde Mobility of TAG-1 Beads on the Surface of NrCAM-transfected B104 Cells
On differentiation in serum-free medium, B104 cells exhibited neurites and growth cone-like projections. We took advantage of the well-defined actin retrograde flow in these motile structures (Lin and Forscher, 1995) to study NrCAM coupling with the actin cytoskeleton. We analyzed the strength of NrCAM anchorage to the cytoskeleton using laser tweezers (Figure 2). TAG-1 beads were trapped and maintained for 5 s at the periphery of lamellipodial structures of differentiated NrCAM-expressing B104 cells. This resulted in stable attachment of the beads to the cell surface (83%, n = 35), and no bead release was observed during the 3-min period after the trap release (Figure 2A). As controls, <20% of TAG-1 beads attached on the parental or GFP-TM–expressing B104 cells (Figure 2A). The major part of the bound TAG-1 beads (80%, n = 65) instantaneously adopted a rearward movement on the lamellipodia away from the leading edge with a mean directional velocity of 2.4 ± 0.4 μm/min (n = 9; Figure 2, B and C; Table 1). The MSD vs. time followed a parabolic curve characteristic of a unidirectional mobility and plotting of X and Y coordinates showed a uniform bead movement over time (Figure 2C). A similar velocity has been recorded for NrCAM on the growth cones of cerebellar granule cells in our previous time-lapse study (Faivre-Sarrailh et al., 1999), and for other CAMs (Kamiguchi and Yoshihara, 2001; Lambert et al., 2002) and corresponds to the velocity of the F-actin rearward flow. To directly demonstrate the implication of F-actin in the rearward movement, cells were treated with 2 μM cytochalasin D, a drug which depolymerizes actin filaments. This treatment abolished the retrograde mobility of most TAG-1 beads positioned on NrCAM-expressing cells. Instead, the beads adopted a stationary behavior (86%, n = 23) and the MSD curves indicated a reduced mean directional velocity of 0.5 ± 0.2 μm/min (n = 4; Figure 2, B and D). Such stationary behavior of NrCAM receptors may depend on their association with a cortical actin cytoskeleton resistant to depolymerization. To test this hypothesis, the laser trap was placed ∼0.5 μm behind the bead (where the trap force is maximal) trying to displace the bead. Indeed, most TAG-1 beads were resistant to displacement by the laser trap in the presence of cytochalasin D (77%, n = 21; Table 1), suggesting that the receptors were anchored to some static elements.
No rearward mobility
| % Escaped initial trap
| % Escaped retrapping
| % Retrapped
| % Escaped retrapping
| % Retrapped
| Mean coupling index
|1.80 ± 0.12
|1.56 ± 0.15
|1.88 ± 0.18
|0.78 ± 0.18b
|0.67 ± 0.13b
|0.84 ± 0.11b
|0.82 ± 0.10b
|1.01 ± 0.14b
| 1.75 ± 0.25
Deletion of the Cytoplasmic Tail of NrCAM Does Not Abrogate the Retrograde Mobility of TAG-1 Beads
We postulated that the coupling of NrCAM to the actin retrograde flow may rely on its cytoplasmic domain because several motifs of interaction with the cytoskeleton have been identified (Figure 1A). To analyze the role of the cytoplasmic tail, a truncated form of NrCAM (NrCAMΔcyt) with only two residues remaining in the cytoplasmic domain was stably expressed in B104 cells. Selected clones expressing the same amount of full-length and deleted NrCAM were used in laser trapping experiments (Figure 1B). TAG-1 beads attached well to NrCAMΔcyt-expressing cells (95%, n = 84). Most of the beads (70%, n = 80) moved rearward with a velocity of 1.5 ± 0.7 μm/min (n = 7) and rapidly became resistant to displacement with the laser trap (84%, n = 80; Figure 3A, Table 1), a behavior similar to that on cells expressing full-length NrCAM. Thus, deletion of the cytoplasmic tail did not impair the retrograde mobility of NrCAM and its tight coupling to the cytoskeleton.
Linkage with the Cytoplasmic Tail and cis-interaction through the FNIII Domains Independently Drive NrCAM Retrograde Mobility on the Cell Surface
Next, we made the hypothesis that NrCAM may associate in cis with transmembrane proteins via its extracellular domain when clustered by TAG-1 beads. Binding experiments with soluble NrCAMFc preclustered with anti-Fc indicated that NrCAM binding sites were indeed present at the surface of B104 cells (unpublished data). Such cis-interacting molecules may provide an indirect linkage between NrCAM and the cytoskeleton. We hypothesized that NrCAM may be engaged in a cis-interaction via its FNIII domains, triggered by trans-interaction of the Ig domains with TAG-1 beads. To test this hypothesis, we generated a NrCAM construct in which the four FNIII domains were replaced by GFP (NrCAMΔfn; Figure 1A). As shown in Figure 3A, deletion of the FNIII domains did not impair NrCAM anchorage to the cytoskeleton. Indeed the major part of TAG-1 beads (72%, n = 36) moved rearward shortly after their attachment to the cell surface and could not be recaptured by the laser trap (Figure 3A, Table 1). Thus, FNIII domains seemed to be dispensable for the coupling of NrCAM to the retrograde actin flow.
We then generated a construct with deletions of both the FNIII domains and the cytoplasmic tail (NrCAMΔfnΔcyt). This construct was stably transfected into B104 cells and two clones were selected on the basis of their strong receptor expression at the cell surface as judged by Western blotting (Figure 1B). TAG-1 bead binding was observed to occur on NrCAMΔfnΔcyt-expressing cells (97%, n = 82) as on cells transfected with full-length NrCAM. In contrast with their behavior on clones expressing NrCAM constructs deleted of only their cytoplasmic tail or their FNIII domains, TAG-1 beads rarely moved rearward on NrCAMΔfnΔcyt-expressing clones (Figure 3). The percentage of beads displaying retrograde mobility was strongly reduced to 27% (vs. 80% in cells expressing full-length NrCAM; Table 1). It should be noted that the low percentage of TAG-1 beads that moved rearward adopted an irregular trajectory exhibiting temporary diffusion periods (Figure 3C). In addition, the percentage of TAG-1 beads that escaped to laser retrapping strongly decreased (33 vs. 80% for full-length NrCAM; Figure 3A, Table 1). Therefore the NrCAMΔfnΔcyt construct exhibited a reduced capability of anchorage to the cytoskeleton.
Surprisingly, only few of the nonrearward moving TAG-1 beads freely diffused on NrCAMΔfnΔcyt-expressing cells (27%, n = 30; Figure 3, A and D), as observed for TAG-1 beads that unspecifically attached on parental B104 cells. The major part of the nonrearward moving beads remained stationary during the 3-min recording period (50%, n = 30), as if they were confined to a specific domain of the lamellipodia (Figure 3B). Beads that were judged immobile by eye exhibited either a small directional movement or a complete immobility, resulting in a very low mean velocity 0.1 ± 0.2 μm/min (n = 9). However, despite this reduced mobility, a large percentage (64%) of the nonrearward-moving beads were displaced by the laser trap, indicating loose anchoring to the cytoskeleton.
A coupling index was attributed to the different NrCAM constructs, based on three parameters: escaping initial trap before the 5-s period of bead attachment, rearward movement, and resistance to retrapping after bead attachment (Table 1). A statistical analysis was carried out to compare the coupling index of the different NrCAM constructs. Single deletions of the cytoplasmic tail or FNIII domains did not result in any significant alteration of the coupling with the cytoskeleton, whereas deletion of both regions had a significant effect (Table 1). Several conclusions can be drawn from the results with the NrCAMΔfnΔcyt construct. First, we can infer that in absence of the FNIII domains, the cytoplasmic tail is required for a direct coupling of NrCAM with the actin retrograde flow. Second, we can conclude that in absence of the cytoplasmic tail of NrCAM, the FNIII domains are implicated in the recruitment of NrCAM into a heterocomplex with a cis partner that provides an efficient anchorage to the actin cytoskeleton. Taken together, these data point to a role for the cytoplasmic tail of NrCAM in its coupling with the actin retrograde mobility, which should be reinforced by cis-association mediated by the FNIII region.
Role of the Cytoplasmic Domain of NrCAM in Its Coupling with the Actin Cytoskeleton
Next, we initiated mapping of the cytoplasmic region of NrCAM involved in the coupling with the F-actin retrograde flow. NrCAM is subject to extensive alternative splicing in six different regions of the extracellular domain, but also in the cytoplasmic region. Interestingly, a splicing variant has been characterized from human and chicken brain that displays a stop codon before the last 82 amino acids (Dry et al., 2001). If translated, this mRNA would generate a NrCAM isoform lacking the highly conserved ankyrin-binding domain and the C-terminal PDZ-domain binding motif. We generated a NrCAM construct with a truncation of the last C-terminal 70 amino acids (NrCAMΔfnΔCter) that was stably transfected in B104 cells (Figure 1, A and B). Most of TAG-1 beads attached to the cell surface but did not display rearward movement (Table 1, Figure 3B), as also observed for the NrCAMΔfnΔcyt construct. This stationary behavior resulted in a low velocity of 0.2 ± 0.1 μm/min (n = 4). Thus, the C-terminal region is critically involved in receptor anchorage with the retrograde actin flow. However, the percentage of TAG-1 beads that resisted to displacement by the laser trap was estimated to 72 vs. 33% for the NrCAMΔfnΔcyt construct (Figure 3A), indicating that the 44 aa juxta-membrane region of the cytoplasmic tail may be implicated in interaction with the cytoskeleton. The juxta-membrane region is involved in the colocalization of L1 with actin stress fibers in transfected glioma cells and was proposed to potentially mediate interaction with α-actinin (Dahlin-Huppe et al., 1997). In addition, the juxta-membrane region contains a consensus motif for the binding of FERM proteins (Hoover and Bryant, 2000).
Because both α-actinin and the FERM protein ezrin are present in the peripheral domain of lamellipodia, we examined the possibility of NrCAM association with these actinbinding proteins. Using long-term binding assays with 4-μm TAG-1 beads on NrCAM-expressing cells, we analyzed the distribution of both proteins and actin filaments by immunofluorescence. We observed that F-actin (Figure 4A) and α-actinin (Figure 4B) were both heavily recruited and colocalized at the site of bead-cell contact. In contrast, the ezrinrich filopodial structures were found apposed to the TAG-1 beads, but ezrin did not appear to be recruited at the contact site (Figure 4C). The recruitment of α-actinin was observed at the contact for 36% (n = 209) of the beads attached to the lamellipodia of NrCAM-expressing cells. In addition, we tested the recruitment of other adaptors involved in the cytoskeletal linkage at adherent contacts. Talin, which has been implicated in the mechanical coupling between integrins and actin filaments (Giannone et al., 2003), was not observed to be recruited under TAG-1 beads.
NrCAM Partitioning with Lipid Rafts
Laser-trapping experiments with TAG-1 beads indicated that clustered NrCAMΔfnΔcyt receptors displayed very reduced mobility in the absence of interactions with the cytoskeleton, suggesting that they were associated with membrane structures supporting confinement. We hypothesized that such reduced mobility might be due to the recruitment of clustered receptors into lipid rafts (Pralle et al., 2000; Suzuki et al., 2000; Tanimura et al., 2003). Indeed, NrCAM contains a palmitoylation sequence in its trans-membrane domain that is conserved in the L1-family, and neurofascin has been reported to be palmitoylated at this site and to partition with lipid rafts (Ren and Bennett, 1998). Because NrCAM is Triton X-100 soluble in microsomes, we used Lubrol to test for the partitioning of NrCAM into detergent-resistant membranes. NrCAM- and NrCAMΔfnΔcyt-expressing cells were extracted with Lubrol and subjected to an equilibrium sucrose gradient centrifugation. Both molecules were present in the low-density lipid raft fraction and copartitioned with the raft component caveolin-1 (Figure 5A). In contrast, the cytoskeleton linker α-actinin was only recovered in the high-density fractions at the bottom of the gradient (Figure 5A). The partitioning of NrCAM into the lipid raft fraction was mediated by its transmembrane domain because the membrane-spanning domain of NrCAM alone fused with GFP as an extracellular region (GFP-TM) was also present in the raft fractions (Figure 5A).
We then analyzed the distribution of NrCAM after antibody- or TAG-1–mediated clustering at the cell surface and its resistance to Triton X-100 extraction in living cells. Both full-length NrCAM (unpublished data) and NrCAMΔfnΔcyt were detected in clusters that were resistant to Triton X-100 extraction (Figure 5, B, D and F). A treatment with MBCD to deplete membrane cholesterol did not modify cell surface expression (Figure 5C) but completely altered the resistance to Triton X-100 extraction of NrCAMΔfnΔcyt (Figure 5, E and G).
To further confirm the association of NrCAM with lipid rafts, coimmunofluorescence staining of NrCAM was carried out with filipin, a fluorescent molecule that is used to detect free cholesterol. As shown in Figure 5H, NrCAM, which mediated homophilic adhesion was highly concentrated at cell-cell contacts (arrowheads). Strikingly, intense filipin fluorescence was colocalized with NrCAM labeling at these contact sites (Figure 5I).
To determine whether the confined behavior of TAG-1 beads may result from the recruitment of NrCAMΔfnΔcyt into lipid rafts, we performed immunofluorescence stainings for caveolin-1. Caveolin-1 strongly accumulated at the contact site with TAG-1 beads (Figure 5, J and L). Quantitative analysis indicated that caveolin-1 was recruited under 32% of TAG-1 beads (n = 80). In contrast, caveolin-1 did not accumulate under TAG-1 beads nonspecifically bound to GFP-TM–expressing B104 cells (Figure 5, K and M). As a control, anti-Fc–coated beads very seldom bound to the cells and recruited neither NrCAMΔfnΔcyt nor caveolin-1 (unpublished data). This result demonstrates that a pool of raft-associated NrCAM is recruited by the binding of TAG-1 beads.
Lipid Raft Association Is Involved in NrCAM Coupling to the Cytoskeleton
MBCD treatment was further used to analyze the consequence of raft dispersion on TAG-1 bead attachment and rearward transport using laser tweezers experiments. B104 cells expressing full-length NrCAM or NrCAMΔfn were treated with MBCD for 15 min before positioning TAG-1 beads with the laser trap. The drug caused partial retraction of some cells, so we were careful to select only cells showing normal lamellipodial motility. The beads attached as in control conditions (100%) but only 30% displayed a rearward movement. Instead, most beads remained confined in the peripheral region of the lamellipodia and were resistant to displacement with the laser trap (62%, n = 16; Table 1, Figure 6, A and C). These confined beads had a mean velocity of 0.5 ± 0.2 μm/min (n = 6). These results indicate that NrCAM is able to interact with immobile components of the cytoskeleton in a way independent from its partitioning with lipid rafts, but that its anchorage to the retrograde actin flow critically requires lipid raft integrity. Next, we analyzed the effect of MBCD treatment on cells expressing NrCAMΔfnΔcyt. TAG-1 beads that were maintained in contact with the cell surface during 5 s did not attach well and the low percentage of attached beads (35%) underwent long-range diffusion (85%, n = 7; Figure 6A).
Because the specificity of MBCD has been recently questioned (Edidin, 2003; Kwik et al., 2003), we carried out a series of control experiments. First, MBCD saturated with exogenous cholesterol did not change the behavior of TAG-1 beads compared with that of untreated cells (Figure 6, A and D). Under this condition, TAG-1 beads displayed a rearward movement (78%, n = 32) with a mean velocity of 2.7 ± 0.3 μm/min (n = 10). Thus, the transition from retrograde to stationary behavior induced by MBCD was specific of cholesterol depletion. Second, because a recent study indicates that MBCD treatment perturbs the phosphoinositide metabolism, which in turn influences actin organization (Kwik et al., 2003), control experiments were performed using neomycin sulfate to sequester membrane PI(4,5)P2. Cells were treated with 10 mM neomycin sulfate for 8 h because a longer treatment was toxic. After neomycin treatment, most TAG-1 beads displayed a rearward mobility with a mean velocity of 2.1 ± 0.3 μm/min (n = 8, Figure 6, A and E). Thus, neomycin did not block the rearward mobility of NrCAM. Third, beads coated with con-A were used to evaluate global effects of MBCD on lamellipodial actin flow (Suter et al., 1998). To make sure that con-A would not attach to NrCAM, we used the clone GFP-TM. Con-A beads attached and displayed rearward mobility with a mean velocity of 4.3 ± 1.1 μm/min (n = 4, Figure 6F). After MBCD treatment, most con-A beads (76%, n = 17) still moved rearward, albeit with a slower velocity of 1.5 ± 0.7 μm/min (n = 5, Figure 6G). This indicates that MBCD reduces the speed of the cortical actin flow but does not affect the coupling of lectin receptors to it.
Lipid Raft Association Is Involved in Ligand Binding and Clustering of NrCAM
In a second step, we analyzed TAG-1 bead binding in long-term assays in the presence of MBCD. Cells were first treated with 10 mM MBCD for 30 min and incubated with 4-μm TAG-1 beads for 30 min. We observed 72 and 76% reduction of bead binding on cells expressing NrCAMΔcyt and NrCAMΔfnΔcyt, respectively, after MBCD treatment. Binding of TAG-1 beads on full-length NrCAM-expressing cells was reduced by 53% in contrast with laser trapping assays in which forced interaction resulted in attachment of the beads in most of the trials. Because MBCD treatment did not reduce cell surface expression of NrCAM (Figure 5C), we hypothesize that under this experimental condition, lipid raft dispersion precluded the clustering of receptors, thus preventing efficient binding of the ligand.
To address this point, a quantitative analysis was performed to evaluate receptor clustering under control condition and after a brief MBCD treatment (Figure 7). Cells were first incubated with 4-μm TAG-1 beads for 15 min and then treated or not with 10 mM MBCD for 15 min. Under this experimental condition, the mean number of beads attached to the cell surface was not modified by MBCD treatment. The GFP-fluorescent clusters of NrCAMΔfnΔcyt at the beadcell contacts displayed a mean diameter of 1.8 ± 0.4 μm (Figure 7, A, B, and F). After 15-min treatment with MBCD, this diameter was reduced to 1.2 ± 0.3 μm (Figure 7, C, D, and F) and clusters of receptors were only detected under 28% of TAG-1 beads (vs. 62% under control condition; Figure 5E). MBCD treatment also altered the recruitment of α-actinin at the contact with TAG-1 beads in NrCAM-expressing cells. Alpha-actinin was only detected under 13% of TAG-1 beads (n = 173), compared with 36% under control condition. Therefore, disruption of rafts by MBCD altered the clustering of NrCAM molecules and the recruitment of α-actinin under TAG-1 beads that may in turn impair anchoring to the retrograde actin flow.
Role of the Cytoplasmic Domain and Raft-Partitioning on the Mobility of NrCAM
Finally, we analyzed the effect of bead clustering and MBCD treatment on the lateral mobility of NrCAM, using FRAP on GFP-tagged NrCAMΔfn and NrCAMΔfnΔcyt clones. Cells were incubated with TAG-1 beads and then treated or not with MBCD. A single photobleach was applied in areas of contact with TAG-1 beads or on control lamellipodia, and the fluorescence recovery was monitored in the bleached spot (Figure 8, A and B). In control areas with unclustered receptors, NrCAMΔfn fluorescence reached a plateau of 54% of the prebleach value with a characteristic recovery time around 20 s (Figure 8C, n = 8). The NrCAMΔfnΔcyt construct exhibited a similar recovery time and a plateau value of 68% (Figure 8D, n = 5). Thus, deletion of the cytoplasmic tail mediating anchorage with the cytoskeleton slightly increased the mobile fraction of NrCAM. Treatment with MBCD had no effect for either construct, suggesting that lipid raft disruption did not affect the mobility of unclustered receptors. The recovery of NrCAM fluorescence was then monitored in regions of contact with TAG-1–coated beads, where receptors accumulated (Figure 8B). In these patched areas, the mobile fractions of NrCAMΔfn and NrCAMΔfnΔcyt were significantly reduced by almost a factor of two, indicating that receptors were being trapped by long-lasting interactions with TAG-1 ligands (Table 2). The recovery time of NrCAMΔfn in bead contacts was similar to that in control regions, indicating that the diffusion of mobile receptors was unaltered in these adhesive zones. Also, MBCD treatment did not affect the recovery time of NrCAMΔfn in bead areas, indicating that the mobility of the full-length protein was not sensitive to raft structure. Deletion of the cytoplasmic tail slightly increased the mobile fraction of NrCAMΔfnΔcyt in the area of bead contact, as observed for the unclustered receptor. Interestingly, the recovery time was increased at the bead contact, showing that the unbound NrCAMΔfnΔcyt under the bead has a reduced mobility due to a bead-mediated clustering of the receptor. This could be due to either free NrCAM corecruitment with the ligand-engaged receptors in merged rafts or to transient interactions of unbound NrCAM with the raft. Indeed MBCD treatment significantly decreased this recovery time to the value of the unclustered receptor (Figure 8D, Table 2). All together, these results suggest that when NrCAM receptors are clustered upon TAG-1 binding, both the coalescence of rafts and anchorage with the cytoskeleton via the cytoplasmic tail are induced and result in lower receptor mobile fraction and/or diffusion coefficient.
Recovery fraction (%)
Recovery time (s)
|54 ± 6
|16 ± 1
|52 ± 4
|21 ± 1
|25 ± 3
|16 ± 3
|25 ± 3
|22 ± 3
|68 ± 2
|18 ± 3
|65 ± 5
|14 ± 1
|38 ± 6
|39 ± 6
| 39 ± 4
| 21 ± 3
The control of growth cone motility by cell-cell contacts is based on the assembly of adhesion complexes and the coupling of membrane receptors with the underlying actin cytoskeleton, to generate driving forces. Using optical tweezers and single particle tracking of beads coated with TAG-1 ligands, we have analyzed the mobility of NrCAM and its coupling to the actin cytoskeleton in neuroblastoma cells. We present evidence that several stages of interactions are implicated in the linkage of NrCAM to the retrograde actin flow. First, cis-interaction with the FNIII domains of NrCAM and association with the cytoplasmic tail independently lead to anchoring with the F-actin retrograde flow; second, the recruitment of NrCAM into lipid rafts is required for ligand binding and rearward movement of NrCAM. We show that several mechanisms for NrCAM clustering may cooperate: trans-interaction with the ligand is favored by raft partitioning and reciprocally, ligand binding induces raft merging.
Complex Interactions via the FNIII Domains and the Cytoplasmic Tail Mediate Coupling of NrCAM with the F-actin Retrograde Flow
Using TAG-1 beads, we induced both ligand occupancy and receptor clustering resulting in anchoring of NrCAM with the retrograde actin flow. TAG-1 beads escaped the trap during the attachment period with no possible retrapping, reminiscent of N-cadherin–coated beads on N-cadherin–expressing myogenic cells (Lambert et al., 2002). This behavior is indicative of a strong and immediate coupling to the cytoskeleton, at variance with that of integrins showing progressive reinforcement of their linkage to the cytoskeleton when force is applied to them via fibronectin-coated beads (Choquet et al., 1997). This suggests that the various families of adhesion receptors exhibit different physical coupling to the cytoskeleton, in agreement with their different cytosolic adaptors and transductory pathways.
We observed that deletion of the single cytoplasmic tail was not sufficient to abrogate the rearward movement of NrCAM triggered by TAG-1 beads. Our results indicate that TAG-1 binding may induce the recruitment of a transmembrane adaptor via a cis-interaction mediated by the FNIII domains of NrCAM. It is of particular interest to note that the third FNIII domain of the related L1-molecule homomultimerizes and supports RGD-independent interactions with several integrins (Silletti et al., 2000). Additional work is needed to determine whether integrins may act as adaptors implicated in the coupling of NrCAM to the actin cytoskeleton.
On the other hand, deletion of the FNIII domains unmasked the role of the cytoplasmic tail in the rearward movement of the molecule and allowed to initiate mapping of the cytoplasmic motifs involved. Deletion of the last 70 aa at the C-terminus suppressed the receptor coupling with the F-actin retrograde flow, but not its anchoring to the cytoskeleton. We can conclude that the juxta-membrane region at least supports interaction with static components of the cytoskeleton and that the C-terminal region comprising the FIGQY ankyrin-binding site and SFV PDZ-protein binding sequence is required for the coupling with the retrograde actin flow (Figure 9). It has been reported that ankyrin binding on the FIGQY motif of L1 mediates rearward movement during initiation of neurites (Nishimura et al., 2003). In another study, however, ankyrin binding was shown to induce rather a stationary behavior (Gil et al., 2003). Alternatively, the C-terminal PDZ-binding domain may participate to the coupling of NrCAM with the retrograde actin flow.
In addition, we showed that α-actinin is recruited together with F-actin at the cell contact with TAG-1 beads, whereas ezrin and talin are not. Because α-actinin is implicated in the cross-linking of actin filaments and in their anchorage at focal contacts (Liu et al., 2000), it may participate to the linkage of NrCAM with F-actin.
Role of Lipid Rafts in NrCAM Reduced Mobility
We showed that NrCAM, which contains a palmitoylation motif in its membrane-spanning domain, partitions into lipid rafts. Lipid rafts are probably very small (50 nm or below) and dynamic entities that are difficult to observe individually (Pralle et al., 2000; Edidin, 2003). Only when their structural components are clustered can they be identified by light microscopy. Here, we showed by filipin staining the colocalization of NrCAM with cholesterol-rich domains at cell-cell contacts and the recruitment of the raft component caveolin-1 at the adhesive contact between the cell surface and TAG-1 beads. These data suggest that cross-linking of NrCAM by trans-interaction with TAG-1 induces the formation of patches enriched in lipid rafts that can be visualized by immunofluorescence. Reciprocally, we showed that raft-association of NrCAM is crucial for its clustering upon ligand binding. Cholesterol depletion induced by MBCD strongly reduced TAG-1 bead binding on NrCAM-expressing cells, indicating that aggregation of NrCAM mediated by lipid rafts may increase the receptor avidity for its ligand. Organization of small receptor aggregates before ligand binding and rapid coalescence of individual rafts into large rafts upon ligand binding could account for the raft promotion of receptor clustering (Harris and Siu, 2002).
Upon clustering with TAG-1 beads, the NrCAMΔfnΔcyt receptors were not anchored to the F-actin retrograde flow and could be displaced by the laser trap, indicating a weak coupling to the cytoskeleton. However, the truncated receptors exhibited a stationary behavior in the cell membrane that was altered by MBCD treatment (Figure 9). Indeed, raft-associated proteins undergo abnormal diffusion, often exhibiting transient confinement (Sheets et al., 1997; Simson et al., 1998). In agreement with our results, microbeads attached to GPI-anchored proteins have a reduced diffusion coefficient, highly dependent on ligand density and area of the bead contact (Suzuki et al., 2000). Other experiments using FRAP suggested that the decreased mobility of NrCAMΔfnΔcyt receptors observed in clusters was dependent on raft organization. Such mobility changes correlated with raft clustering have been observed for transmembrane proteins such HA (Shvartsman et al., 2003) or the palmitoylated LAT adaptor upon T-cell activation (Tanimura et al., 2003). Our data also illustrate that there seems to be a limited exchange of NrCAM between stabilized rafts under the bead and the surrounding plasma membrane, as inferred by local measurements of membrane viscosity using small beads coupled to raft-associated receptors (Pralle et al., 2000).
Raft Partitioning Is Required for the Coupling of NrCAM with the Cytoskeleton
We have shown that raft dispersion using MBCD treatment abrogates NrCAM translocation with the F-actin retrograde flow. This might be due to MBCD-induced global modification of membrane properties and cytoskeletal organization. First, we checked that membrane expression of NrCAM was not changed after MBCD treatment in contrast to the diminution of several raft-associated proteins after long lasting LDL depletion (Sheets et al., 1997). Second, MBCD treatment has been shown to decrease the lateral mobility of membrane receptors in FRAP analyses (Kenworthy et al., 2004), which was correlated to a PIP2-dependent cortical redistribution of F-actin (Kwik et al., 2003). In contrast, our FRAP analyses showed no diminution in the mobility of unclustered NrCAM after MBCD treatment. In addition, NrCAM coupling with the F-actin retrograde flow was unaffected by neomycin, a drug that alters PIP2 metabolism. Moreover, because con-A beads still exhibited significant rearward movement after MBCD treatment, we assume that alteration of F-actin dynamics could not account for our results. Thus, the effect of MBCD was assessed to be specific because several drawbacks of cholesterol depletion protocols can be excluded in our study.
A high density of ligand coated on the bead surface is required for generating rearward movement of different types of receptors (Thoumine and Meister, 2000; Coussen et al., 2002; Lambert et al., 2002; Gil et al., 2003). Beads inducing strong cross-linking exhibit a robust coupling with the actin cytoskeleton, and instead beads having lower cross-linking activity attach to the cell surface, but display loose cytoskeletal interactions. Reciprocally, we can assume that the raft-mediated packing of receptors is required for their strong coupling with the actin cytoskeleton (Figure 9). Indeed, we showed that MBCD-induced raft disruption alters NrCAM clustering and α-actinin recruitment at the contact with TAG-1 beads.
Alternatively, raft partitioning should enhance NrCAM-cytoskeleton interactions by selectively concentrating adaptors and making them readily available. Formation of enlarged rafts after cell aggregate formation has been shown to induce specific interactions with the cytoskeleton and to activate specific signaling pathways in Dyctostelium or at the immunological synapse (Moran and Miceli, 1998; Harris and Siu, 2002). For example, recruitment into lipid rafts can induce association with the cortical actin cytoskeleton mediated by caveolin-1 or annexin II (Oliferenko et al., 1999; Kanzaki and Pessin, 2002). In addition, rafts that consist in highly organized signaling platforms could be important for the F-actin clutching at adhesive contacts because this process requires highly regulated phophorylation-based signaling (Suter and Forscher, 2001; von Wichert et al., 2003).
Finally, raft partitioning of NrCAM, which is implicated both in ligand-binding and rearward movement of the receptor, may be critical for the control of axonal growth. Indeed, recent studies indicate that the recruitment of CAMs into lipid rafts is involved in growth cone motility and axogenesis. For example, raft-exclusion of NCAM140 by mutation of its palmitoylation site impairs homophilic-mediated neurite outgrowth (Niethammer et al., 2002). Furthermore, micro-CALI-mediated disruption of rafts slows the L1-dependent migration of DRG sensory neurons (Nakai and Kamiguchi, 2002). Interestingly, disruption of rafts in the peripheral domain, but not in the central domain affects growth cone migration. The peripheral domain is the region where clustering of the L1-molecule triggers anchoring with the F-actin retrograde flow to generate a tension force pulling the growth cone forward, and we demonstrated here that this process requires the integrity of rafts. More generally, it seems that lipid rafts play a pivotal role in the control of axonal guidance because raft-dependent signaling also mediates growth cone turning in response to chemotropic factors (Guirland et al., 2004).
Multiple direct as well as indirect interactions of NrCAM with the actin cytoskeleton may provide a range of modulations of receptor mobility and ability to transduce mechanical forces. Having different possibilities of switching between static adhesion and dynamic coupling may be a key factor to allow a fine control of growth cone motility by addition of different regulatory interactions. Moreover, cis-interaction and raft association would probably allow integration of different environmental cues and coordination of several adhesive events.
Article published online ahead of print. Mol. Biol. Cell 10.1091/mbc.E04–03–0171. Article and publication date are available at www.molbiolcell.org/cgi/doi/10.1091/mbc.E04–03–0171.
Monitoring Editor: Richard Hynes
We thank André Le Bivic, Paul Mangeat, Marianne Renner, and Francis Chaouloff for helpful discussions. This work was supported by the Association pour la Recherche sur la Sclérose en Plaques, the Mizutani Foundation for Glycosciences, the Fondation pour la Recherche Médicale (C.F.-S.), and the Conseil Régional d'Aquitaine (D.C.).