The adhesion molecule N-cadherin plays important roles in the development of the nervous system, in particular by stimulating axon outgrowth, but the molecular mechanisms underlying this effect are mostly unknown. One possibility, the so-called “molecular clutch” model, could involve a direct mechanical linkage between N-cadherin adhesion at the membrane and intracellular actin-based motility within neuronal growth cones. Using live imaging of primary rat hippocampal neurons plated on N-cadherin-coated substrates and optical trapping of N-cadherin-coated microspheres, we demonstrate here a strong correlation between growth cone velocity and the mechanical coupling between ligand-bound N-cadherin receptors and the retrograde actin flow. This relationship holds by varying ligand density and expressing mutated N-cadherin receptors or small interfering RNAs to perturb binding to catenins. By restraining microsphere motion using optical tweezers or a microneedle, we further show slippage of cadherin–cytoskeleton bonds at low forces, and, at higher forces, local actin accumulation, which strengthens nascent N-cadherin contacts. Together, these data support a direct transmission of actin-based traction forces to N-cadherin adhesions, through catenin partners, driving growth cone advance and neurite extension.
- axon elongation
- N-cadherin-Fc fusion protein
- optical tweezers
- single particle tracking
Growth cones are motile structures at the distal extremity of axons responsible for pathfinding and neurite extension during nervous system development and repair. Growth cone motility relies on a dynamic regulation of the actin network, with polymerization occurring at the leading edge, depolymerization in the central region, and the activity of myosins pulling on lamellipodial actin filaments (Lin et al., 1996; Mallavarapu and Mitchison, 1999; Diefenbach et al., 2002). This results altogether in a continuous retrograde flow of actin.
Growth cones adhere to the extracellular matrix and adjacent cells through specific adhesion receptors, i.e., integrins, Ig cell adhesion molecules (IgCAMs), and cadherins. An important question is how growth cone progression on adhesive substrates is coupled to actin motility. A prevalent hypothesis, the “molecular clutch” mechanism, is that the mechanical coupling between ligand-bound adhesion receptors and the actin flow allows traction forces to be transmitted to the substrate, resulting in local diminution of the retrograde flow and forward progression (Mitchison and Kirschner, 1988; Suter and Forscher, 1998). In the case of Aplysia growth cones interacting through ApCAM [the homolog of vertebrate neural cell adhesion molecule (NCAM)], such substrate–cytoskeletal coupling is indeed accompanied by protrusion of the microtubule-rich central domain toward stiff contacts and forward expansion of the actin-rich lamellipodium (Lin and Forscher, 1995; Suter et al., 1998). However, the relationship between the actual translocation of growth cones leading to axonal elongation and the degree of receptor coupling to the actin flow has not been demonstrated.
The adhesion receptor N-cadherin (Ncad) plays a role in a variety of processes in the CNS of vertebrates, including cell positioning (Kadowaki et al., 2007), axon outgrowth (Riehl et al., 1996), fasciculation and dendritic branching (Yu and Malenka, 2003; Bekirov et al., 2007), synaptogenesis (Benson and Tanaka, 1998), and synaptic plasticity (Bozdagi et al., 2000; Togashi et al., 2002; Okamura et al., 2004). The effect of N-cadherin in promoting axonal growth can be mimicked in cell culture models (Matsunaga et al., 1988; Lemmon et al., 1992; Saffell et al., 1997; Kamiguchi and Yoshihara, 2001), but the underlying molecular mechanisms remain unclear. In particular, whether a clutch process exists for N-cadherin-mediated growth cone locomotion and which molecular partners it may involve is still unknown. Here, using primary neurons interacting with N-cadherin-coated substrates together with live imaging and optical manipulation, we show that the mechanical coupling between N-cadherin receptors and the actin flow through catenins is a major determinant of growth cone motility and neurite extension.
Materials and Methods
Ncad-Fc and cadherin 11 (cad11)-Fc were produced at 200 μg/ml as described previously (Lambert et al., 2000; Marthiens et al., 2005). Synaptic CAM (SynCAM)-Fc was constructed from a SynCAM plasmid given by T. Biederer (Yale University, New Haven, CT) (Biederer et al., 2002) and purified similarly (Breillat et al., 2007). GC4 monoclonal antibody against chicken N-cadherin, globulin free BSA, cytochalasin D, and other chemicals were purchased from Sigma.
Plasmids coding for C-terminal green fluorescent protein (GFP)-tagged chicken wild-type N-cadherin (NcadWT) and N-cadherin deleted of the 35 aa C-terminal β-catenin binding region (NcadΔβcat) were described previously (Thoumine et al., 2006). Wild-type chicken N-cadherin fused in C terminus with DsRed tandem dimer (Ncad-DsRed) has been described previously (Lambert et al., 2007). N-terminal myc-tagged Xenopus N-cadherin lacking the extracellular domain (NcadΔextra) was a kind gift from C. Holt (Riehl et al., 1996). Mouse N-cadherin with triple alanine mutation in the juxtamembrane domain known to abolish binding to p120 (NcadAAA) tagged with yellow fluorescent protein (YFP) in the C terminal was a kind gift from K. G. Green (Northwestern University, Chicago, IL) (Chen et al., 2003). Actin–GFP was a gift from A. Matus (Friedrich Miescher Institute, Basel, Switzerland) (Fischer et al., 1998). pEGFP-N1 was from Clontech. To knock down α-catenin expression, we used small interfering RNA (siRNA) duplexes (Proligo) targeting the αE-catenin sequence 5′-AAAGACGTGGATGGGCTGGAT-3 [RNA interference (RNAi)]. Another duplex, 5′-AAGAAGGCCCATGTTTTGGCT-3′, also matching αE-catenin mRNA, but which did not disturb αE-catenin protein expression, was used as a control (RNActr).
Glass coverslips (15 mm) were cleaned overnight with 6 m nitric acid, rinsed with deionized water, and autoclaved. Coverslips were incubated for 2 h at 37°C with 1 mg/ml polylysine, then overnight at 4°C with 4 μg of goat anti-mouse Fc (Jackson ImmunoResearch) in 100 μl of 0.2 m borate buffer, pH 8.5, and finally for 3 h at room temperature with 0.03–3 μg of Ncad-Fc or 3 μg of mouse Fc. Alternatively, substrates were coated with goat anti-human Fc and then incubated with 3 μg of cad11-Fc or 3 μg of SynCAM-Fc. In control experiments, substrates were coated for 2 h with polylysine then incubated overnight at 4°C with 50 μg/ml laminin (BD Biosciences). For optical tweezers experiments, coverslips were treated with polylysine alone.
Cell culture and transfection.
Hippocampal neurons from embryonic day 18 rat embryos were plated on these substrates at a density of 10,000 cells/cm2 in Modified Eagle's Medium (MEM) containing 10% horse serum (Invitrogen) for 3 h, then turned onto a layer of glial cells and cultured in Neurobasal medium supplemented with B27 (Goslin et al., 1991). Cells were transfected using Effectene (Qiagen) or calcium phosphate (Thoumine et al., 2006) with similar results. Cells were manipulated 30–40 h after transfection. In some experiments, 4 mm EGTA, 1:50 GC4 antibody, 1 μm cytochalasin D, or 1:1000 DMSO were added to the culture or observation media. α-Catenin downregulation by RNA interference was performed by cotransfection of GFP vector as a reporter with RNA duplexes, using calcium phosphate or Lipofectamine 2000 (Invitrogen): 3 μl of reagent, 0.6 μg of vector, and 0.9 μg of RNAi or RNActr were used to transfect one 15 mm glass coverslip. Neurons were incubated for 1 h with this solution in MEM supplemented with 2% B27, 2 mm glutamine, 1 mm Na-pyruvate, and 1 m HEPES (Eugene et al., 2007). Forty-eight hours after transfection, siRNA efficiency was checked by Western blot and immunostaining.
Quantification of growth cone velocity and neurite arborization.
Cells at 1–2 d in vitro (DIV) were mounted in Neurobasal medium containing 20 mm HEPES for live observation on an inverted microscope (Olympus IX50) using a 40×, 1.3 numerical aperture (NA) objective and differential interference contrast (DIC) illumination. Images were acquired every 10 s for 30 min using a CCD camera (HQ CoolSnap; Roper Scientific) driven by the MetaMorph software (Universal Imaging). Three regions were recorded simultaneously by scanning the sample with the motorized stage (MarzHauser). Temperature was maintained at 37°C with an air blower (World Precision Instruments) and an objective heater (Bioptechs). The centroid coordinates of individual growth cones, as well as the number and length of neurites from neuronal cultures fixed with 4% paraformaldehyde, were analyzed using NIH ImageJ.
Five microliters of 1 μm sulfate latex microspheres (Polysciences; 4.5 × 107 particles/μl) were incubated overnight at 4°C with 10 μg of goat anti-Fc antibody (Jackson ImmunoResearch) in 50 μl of 0.2 m borate buffer, pH 8.5. Microspheres were rinsed in borate buffer containing 0.3% BSA, and 20 μl of the suspension was incubated at room temperature for 3 h with 10 μg of cad11-Fc, SynCAM-Fc, or Fc, or 0.1–10 μg of Ncad-Fc. Beads were rinsed again three times, resuspended in 100 μl, and kept on ice during the experiments. The same protocol was used to coat 5 μl of 4 μm magnetic microspheres (density, ρm = 1.3 mg/ml; 4 × 105 particles/μl; Dynal), or 2 and 4 μm sulfate microspheres (2.7 × 106 particles/μl; Interfacial Dynamics) with 2 μg Ncad-Fc for bead adhesion and micromanipulation experiments, respectively.
Bead adhesion assays.
Coverslips containing neurons at 2 DIV were placed in a 12-well plate using 1 ml of culture medium supplemented with 1% BSA per coverslip and incubated at 37°C with 1 μl of bead suspension. After 30 min, nonadherent beads were rinsed away using an aspiration flow rate of 1 ml/s. This produces a Stokes force of 6πηav, where η is the medium viscosity (10−3 kg · m−1 · 5−1), a is the radius of the bead, and v the fluid velocity (∼1 cm/s). In the case of magnetic beads, coverslips were turned upside down and the 12-well plate was centrifuged at a relative centrifugal force (RCF) of 1730 × g for 15 min to pull on adherent beads. The force applied was then (4/3)πa3(ρm − ρl) RCF, where ρl (1 mg/ml) is the fluid density. Cells were then fixed in 4% paraformaldehyde/4% sucrose in PBS, mounted on microscope slides, and observed using a 40× objective. In the case of transfected cells, immunocytochemistry was performed as described previously (Thoumine et al., 2006) using 1:1000 anti-GFP rabbit (Invitrogen) or chicken (Millipore) serum, 1:50 monoclonal anti-rat N-cadherin (Transduction Laboratories), 1:500 polyclonal anti-myc (Upstate), or 1:75 anti-αE-catenin rabbit (Epitomics) or mouse (Santa Cruz Biotechnology) serum as primary antibodies, and 1:1000 Alexa-568-conjugated secondary antibodies (Invitrogen). Ratios of intensity of α-catenin and control stainings from GFP-expressing neurons were assessed using ImageJ software.
Optical tweezer experiments.
An inverted microscope was fed through its epifluorescence port by an Nd:YAG laser beam (Compass; Coherent). Trapping was achieved with a laser power of 100 mW at the back plane of a 100×, 1.40 NA objective. Microspheres captured using a motorized stage (MarzHauser) were maintained at the periphery of growth cones for 2 s, then their movement in DIC illumination was recorded for 2 min at 10 Hz. Trajectories were tracked using MetaMorph, and the mean squared displacement over time was computed using a homemade algorithm (Thoumine et al., 2005). Alternatively, the trap was applied continuously for 2 min. The trap stiffness k was estimated by measuring the position fluctuations Δx of captured beads at 40 Hz. Using the equipartition theorem k<Δx2> = kBT, where kBT is the thermal energy, we found k = 4.4 pN/μm. We checked for linearity between laser power and trap stiffness.
Microneedles were pulled from 1-mm-diameter glass rods (Clark Electromedical) using a micropipette puller (Sutter Instruments) and coated with 1% BSA. Microneedle stiffness estimated from the geometric profile of the tip and the knowledge of the glass Young's modulus was in the range of 1 nN/μm. Coarse positioning was achieved using a 20× objective and a three-axes motor-driven micromanipulator (Eppendorf). Fine positioning was performed using a piezoelectric device (Burleigh) and a 100× objective. Growth cones were oriented with respect to a fixed microneedle using a manual rotation plate (Thorlabs). Ncad-Fc-coated beads (4 μm) were maintained on growth cones for 10 s, then left alone or immediately restrained with the microneedle and pulled forward at a velocity of ∼1 μm/min. Alternating transmission and GFP fluorescence images were taken at 10 Hz for 10 min.
For the quantification of neurite length, growth cone velocity, and coupling index, data from three to five separate experiments were pooled and expressed as mean ± SEM. The number of cells or beads examined in each condition is indicated in italics on the graphs in the Figures. All coating conditions or transfections were run in parallel on the same batches of neurons and compared by one-way ANOVA using GraphPad Prism. Individual conditions were compared with controls (Fc or GFP, respectively) by Dunnett's test, or two by two using Tukey's test. Correlations were calculated using a linear fit and Pearson's test.
N-cadherin selectively stimulates neurite extension and growth cone migration
We first developed an experimental system supporting N-cadherin-specific neuritic growth. We used rat embryonic hippocampal neurons plated on glass coverslips coated with purified N-cadherin-Fc. Ncad-Fc induced a more developed neuritic network when compared with substrates coated with other adhesion proteins cad11-Fc or SynCAM-Fc, or Fc alone (Fig. 1A). This neuritogenic effect of N-cadherin was similar to that reported previously for other neurons (Lemmon et al., 1992; Saffell et al., 1997; Kamiguchi and Yoshihara, 2001). The number of neurites per cell (data not shown) and the length of the longest neurite, most likely the axon, increased with Ncad-Fc coating density (Fig. 1D). We also characterized growth cone locomotion using time-lapse microscopy (Fig. 1B; supplemental Videos 1, 2, available at www.jneurosci.org as supplemental material). The displacement of growth cone centroids increased in a steady manner over time, without pauses, allowing the calculation of a long-term velocity (Fig. 1C). Growth cone velocity increased with Ncad-Fc coating density and sharply correlated with neurite length (Fig. 1D), indicating that the neuritogenic effect of N-cadherin is driven by a stimulation of growth cone migration. The linear relationship agrees with the concept that growth cones extend neurites by pulling on them, the velocity of growth cones being translated into a constant rate of neurite extension (Lamoureux et al., 2002). Those effects were specific of N-cadherin adhesion because addition of the calcium chelator EGTA or the GC4 antibody to block N-cadherin homophilic binding drastically reduced growth cone advance and neurite extension (Fig. 1D).
N-cadherin anchorage to the actin flow correlates with growth cone velocity
Next, we characterized the physical coupling between N-cadherin and the underlying actin flow using Ncad-Fc-coated microspheres manipulated by optical tweezers. Neurons were plated on polylysine to avoid mobilization of endogenous N-cadherin receptors at the basal surface. Microspheres were placed on growth cones for 2 s, then either the optical trap was released and bead trajectories were followed for 2 min, or the trap was applied continuously and the latency for beads to escape was quantified (Fig. 2, top). One hundred percent of beads coated with 1 μg/cm2 Ncad-Fc density stayed adherent to the dorsal growth cone surface (n = 40), and that fraction was reduced to 32% in the presence of EGTA (n = 26) or 12% using function-blocking antibodies to N-cadherin (n = 24), indicating adhesion specificity. Microspheres coated with high Ncad-Fc density (0.3 and 1 μg/cm2) showed directed retrograde movement with minimal lateral diffusion (Fig. 2A; supplemental Video 3, available at www.jneurosci.org as supplemental material) and could escape the optical trap in a few seconds (Fig. 2D). At intermediate Ncad-Fc densities (0.1 and 0.03 μg/cm2), bead trajectories were more irregular and characterized by a higher degree of lateral mobility (Fig. 2B). Beads also showed a longer latency period before escaping the trap (Fig. 2E). Microspheres coated with minimal Ncad-Fc densities (0.01 μg/cm2) or Fc alone adhered poorly and showed essentially Brownian motion (Fig. 2C; supplemental Video 4, available at www.jneurosci.org as supplemental material). They also stayed for minutes in the continuously applied trap (Fig. 2F) and could easily be dragged along the neuronal surface (data not shown). For each trajectory, the mean squared displacement over time (t) was fitted by the diffusion/flow equation 4Dt + V2t2 (Fig. 2G), where D and V represent the diffusion coefficient and retrograde velocity of the bead, respectively (Kucik et al., 1991). The velocity of Ncad-Fc-coated beads (∼5 μm/min), close to the retrograde velocity of fluorescently labeled actin measured in DRG neurons (Diefenbach et al., 2002), was independent of Ncad-Fc concentration and dramatically reduced by the addition of cytochalasin D, which abolishes actin dynamics (Table 1). Thus, the parameter V likely reflects the actual speed of the actin retrograde flow. In contrast, the diffusion coefficient D decreased as Ncad-Fc density was raised, reflecting higher coupling (Table 1). We also measured the fraction of beads escaping the trap before 1 min (Pe), which increased smoothly with Ncad-Fc concentration and dropped to zero in the presence of cytochalasin D (Table 1).
We pooled the effects on bead diffusion, velocity, and escape into a single robust parameter called a coupling index, and defined as Pe V log(D0/D) (in micrometers per minute). The logarithmic function stems from the fact that the diffusion coefficient of membrane proteins often spans several orders of magnitude (Kucik et al., 1999; Thoumine et al., 2005) and uses a reference diffusion coefficient D0, the one obtained for Fc-coated beads (0.03 μm2/s). After checking by immunofluorescence that the amounts of Ncad-Fc adsorbed on microspheres and glass coverslips increased similarly with coating concentration (supplemental Fig. 1, available at www.jneurosci.org as supplemental material), we plotted the coupling index versus growth cone velocity (Fig. 2H). We observed a sharp correlation, demonstrating that the degree of N-cadherin receptor anchoring to the retrograde actin flow is a good predictor of growth cone advance. The effect was specific of N-cadherin because two other ligands, cadherin-11 and SynCAM, fell well outside the correlation line (Fig. 2H).
Slippage of bonds between N-cadherin receptors and the actin flow at low forces
The fact that even at maximal Ncad-Fc coating density the velocity of growth cones (∼1 μm/min) remains much smaller than the speed of the actin flow (5 μm/min) implies some slippage between the moving actin network and N-cadherin receptors engaged in adhesive interactions at the basal surface. To investigate this process, we more closely examined the behavior of beads coated with medium Ncad-Fc densities (0.03 and 0.1 μg/cm2) maintained on growth cones with the optical trap applied continuously. We observed many breaking events in which beads started to move rearward for 0.3–0.6 μm, then occasionally snapped back into the trap center in <200 ms (Fig. 3A,B; supplemental Video 5, available at www.jneurosci.org as supplemental material). Sixty-eight such events were recorded from 282 beads tested (breaking frequency, 0.24). The frequency of retrapping events was dramatically increased in the presence of cytochalasin D (7 events from 9 beads; frequency, 0.77) versus the control DMSO condition (2 events from 10 beads; frequency, 0.20 similar to untreated cells), indicating that this phenomenon reflects a transient linkage to the actin network. By multiplying the distance traveled from the trap center by the trap stiffness, we computed the corresponding breaking forces, on average 1.6 ± 0.1 pN (n = 68). The histogram showed a major peak centered at 1.2 pN, together with smaller secondary peaks at 2.5 and 4 pN (Fig. 3C), supporting the notion that bead recapture corresponded to the breakage of individual or very few molecular bonds. Note that 1 μg/cm2 Ncad-Fc beads escape readily (Fig. 3B) because the optical trap is not strong enough compared with the forces generated by the actin flow. We will show later an assay using a microneedle in which we can indeed restrain microspheres and demonstrate slippage at high Ncad-Fc coating density.
Perturbing interactions between N-cadherin receptors and catenins similarly affect neurite outgrowth and receptor–cytoskeletal coupling
To investigate the molecular determinants underlying the anchorage of N-cadherin to the actin flow, we first used an approach based on the expression of mutated N-cadherin receptors. Neurons were transfected with either NcadΔextra (Riehl et al., 1996), NcadΔβcat (Thoumine et al., 2006), NcadAAA (Chen et al., 2003), NcadWT, or GFP as a control. We showed previously that these receptors were expressed approximately twofold over endogenous N-cadherin and correctly addressed to the neuronal surface (Thoumine et al., 2006). We first characterized the effect of these constructs on neurite elongation for neurons plated on Ncad-Fc, and tested their specificity by comparing with cells grown on laminin. Transfection of NcadWT resulted in a slight (18%), but not significant, diminution in neurite length when compared with GFP-positive cells, an effect that was also found for cells plated on laminin (Fig. 4). In contrast, neurons expressing NcadΔextra, NcadΔβcat, and NcadAAA all exhibited significantly shorter neurites compared with the GFP condition, with average reductions in length of 50, 30, and 45%, respectively (Fig. 4). The effects of those mutants on neurite outgrowth were sometimes so strong that the neurons appeared to stay in a rather unpolarized stage, with little axon–dendrite differentiation (Bradke and Dotti, 1997; Lamoureux et al., 2002). The effect of NcadΔextra was highly specific for Ncad-Fc because its expression did not alter axon growth on laminin (Fig. 4B). However, the effects of NcadΔβcat and NcadAAA were less selective, because their expression also reduced neurite extension on laminin by ∼30%. There are several possible explanations for this behavior which will be addressed in the discussion.
Next, the coupling of mutated N-cadherin receptors to the actin flow was estimated by optical tweezers (Fig. 5). The rearward velocity of Ncad-Fc beads placed on cells transfected with GFP or NcadWT was slightly reduced when compared with nontransfected counterparts (compare Tables 1, 2), indicating that the actin flow was perturbed by the transfection protocol itself. However, given considerable slippage, this velocity was still sufficient to allow for growth cone advance at normal speed, leading to normal neurite length for GFP-expressing cells (∼200 μm for an Ncad-Fc coating density of 0.3 μg/cm2) (compare Figs. 1D, 4B). In addition, Ncad-Fc-coated beads rapidly coupled to the actin flow in neurons transfected with GFP or NcadWT even in the presence of a continuous trap (Fig. 5A,B), resulting in high escape probability (Fig. 5C) and coupling index (Fig. 5D), and demonstrating intact N-cadherin anchorage to actin. As for the effect on neurite length, there was a slight and nonsignificant reduction in coupling index for NcadWT, which may be explained by a higher adhesion (see below, Weak correlation between N-cadherin binding strength and growth cone migration, and Discussion). In contrast, Ncad-Fc beads on cells transfected with NcadΔextra, NcadΔβcat, and NcadAAA showed more difficulty to escape when compared with GFP-expressing counterparts (Fig. 5A–C). Together with reduced bead velocity and increased diffusion coefficient (Table 2), this resulted in significantly lower coupling index for N-cadherin mutants compared with the GFP and NcadWT conditions (Fig. 5D). Although here again the effects of NcadAAA and NcadΔβcat may affect the actin flow in a nonspecific manner (see Discussion), the coupling index was overall strongly correlated with axonal length in these conditions of perturbation of cadherin-catenin interactions (Fig. 5D).
To gain further insight into the molecular interface between N-cadherin receptors and the actin flow, we used another strategy based on RNA interference against α-catenin. Because the main α-catenin isoform in cultured hippocampal neurons is αE-catenin (supplemental Fig. 2A, available at www.jneurosci.org as supplemental material), we specifically designed and tested siRNA against αE-catenin, together with control siRNA. Transfection of these siRNA in heterologous cells (supplemental Fig. 2B,C, available at www.jneurosci.org as supplemental material) or neurons (Fig. 6A, supplemental Fig. 2D, available at www.jneurosci.org as supplemental material) yielded a dramatic downregulation of αE-catenin expression compared with control-transfected cells (∼50% measured by immunoblot). Neurons with RNAi against αE-catenin grew significantly shorter neurites on Ncad-Fc-coated substrates than those transfected with control siRNA (Fig. 6A,B). This effect was specific of the N-cadherin adhesive interaction because no significant difference was observed when cells were seeded on polylysine-coated substrates (Fig. 6B). The αE-catenin RNAi also interfered with the ability of Ncad-Fc-coated beads placed on growth cones to escape the optical trap (Fig. 6C,D). Indeed, the escape probability was reduced by 60% with respect to control RNA (Fig. 6E). Thus, reducing specifically α-catenin protein expression reduces both neurite extension and coupling efficiency, strongly suggesting that αE-catenin is a key molecule in the connection between N-cadherin adhesion and the retrograde actin flow.
Stiffening of N-cadherin contacts is associated with local actin recruitment
Despite extensive slippage, growth cones slowly tract themselves on immobilized N-cadherin ligands. In this process, the actin network is likely to generate forces on a small number of sufficiently strong N-cadherin receptor clusters. To examine this issue, we monitored the dynamic behavior of actin and N-cadherin adhesions using dual-color time-lapse imaging of actin-GFP and Ncad-DsRed in growth cones migrating on Ncad-Fc substrates. N-cadherin was rather evenly distributed, but occasionally formed small clusters at the substrate level that could last for 1–2 min (Fig. 7A,B). Actin-GFP often accumulated at these Ncad-DsRed-rich clusters, then disappeared as the clusters dissolved (Fig. 7A,B, supplemental Video 6, available at www.jneurosci.org as supplemental material). These results suggested that the clustering of a few N-cadherin receptors is sufficient to induce local and transient actin accumulation. On average, the level of fluorescence of such clusters normalized by control areas on the same growth cones was 2.08 ± 0.12 for actin-GFP and only 1.36 ± 0.04 for N-cadherin-DsRed (n = 22 clusters from 5 cotransfected cells; p < 0.001 by Student's t test), revealing a threefold difference between N-cadherin and actin enrichment at these contact sites with the Ncad-Fc-coated substrate.
We then investigated whether the mechanical resistance at N-cadherin adhesions was involved in such actin accumulation. We mimicked stiff N-cadherin contacts by imposing forces on microspheres coated with maximal Ncad-Fc density, and monitored actin-GFP redistribution. Because the optical trap was too weak to restrain microsphere motion, we used larger beads and prevented them from moving rearward by positioning a stiff microneedle downstream (Suter et al., 1998), resulting in a continuous slippage of the actin flow beneath the microsphere. The force measured by deflection of the microneedle was ∼1 nN, i.e., the strength of ∼10–20 homophilic bonds between N-cadherins (Perret et al., 2004; Pittet et al., 2008). Thus, slippage in these conditions is likely to involve the complex and random rupture of many receptor–cytoskeleton bonds, impossible to resolve at the individual level as shown in Figure 3. There was no particular enrichment for actin at unrestrained beads as they traveled rearward (Fig. 8A). In contrast, restrained microspheres showed punctual accumulation of actin at the site of force imposition (Fig. 8B, supplemental Video 7, available at www.jneurosci.org as supplemental material), which reached a maximum level in 1–2 min, then stayed rather constant (Fig. 8C). Such actin accumulation displayed kinetics and levels similar to those observed on Ncad-Fc-coated coverslips, suggesting a common molecular mechanism. Similar experiments on cells transfected with Ncad-GFP indicated a much slower and more uniform accumulation of Ncad-GFP receptors around Ncad-Fc-coated beads (Thoumine et al., 2006). Therefore, actin recruitment can be initiated by a moderate accumulation of N-cadherin receptors and is force specific. To confirm that actin accumulation was triggered by a stiff coupling between N-cadherin adhesions and the cytoskeleton, we performed additional micromanipulation experiments on cells coexpressing NcadΔextra to perturb binding between endogenous N-cadherin and the actin flow. In those conditions, only a very transient actin-GFP accumulation was observed at the bead contact in the first 2 min, then disappeared in the long term (Fig. 8C,D, supplemental Video 8, available at www.jneurosci.org as supplemental material). Furthermore, when tension was applied forward with the microneedle, we felt very little resistance from these growth cones, in contrast with control cells transfected with actin-GFP alone, in which growth cones crawled toward the tracted beads (supplemental Videos 7, 8, available at www.jneurosci.org as supplemental material).
Weak correlation between N-cadherin binding strength and growth cone migration
Finally, to determine whether growth cone advance was somehow related to the intrinsic strength of N-cadherin adhesions, we measured N-cadherin homophilic adhesiveness in conditions of varying ligand density and overexpressing mutated receptors. Neurons plated on polylysine were incubated for 30 min with 2 μm Ncad-Fc-coated microspheres and nonadherent beads were rinsed away (Fig. 9A,B) or pulled away by centrifuging the cultures upside down (data not shown). We estimate that such protocols produced mechanical forces of ∼200–250 pN on the beads (see Materials and Methods), the strength of approximately five cadherin molecular bridges (Perret et al., 2004; Pittet et al., 2008). Therefore, this long-term adhesion assay reflects rather strong contacts involving several N-cadherin molecules. The number of beads remaining bound per cell increased gradually with Ncad-Fc density (supplemental Fig. 3A, available at www.jneurosci.org as supplemental material) and was strongly reduced in the presence of EGTA or function-blocking antibodies, indicating adhesion specificity (supplemental Fig. 3B,C, available at www.jneurosci.org as supplemental material). We also probed N-cadherin adhesiveness specifically at growth cones by placing 1 μm microspheres for 2 s at their periphery using optical tweezers (Fig. 2). The beads that diffused away in the bath after initial contact were scored as nonattached. This allowed calculation of the fraction of adherent beads Pa, which measures whether adhesion force is greater than the thermal motion of the bead. This short-term assay showed basically the same trend as the sedimentation assay (Figs. 9, supplemental Fig. 3, available at www.jneurosci.org as supplemental material). As expected, the effect of N-cadherin on growth cone advance and neurite extension was positively related to the degree of N-cadherin homophilic adhesion, but there were a few counter examples. For example, the actin-depolymerizing agent cytochalasin D arrested growth cone progression (Fig. 2H) without affecting N-cadherin adhesiveness (supplemental Fig. 3A,C, available at www.jneurosci.org as supplemental material). Also, overexpression of NcadWT and Ncad-AAA receptors increased bead binding (Fig. 9C) but reduced neurite extension (Fig. 4B). In contrast, neurons transfected with NcadΔextra and NcadΔβcat showed a modest reduction in adhesiveness (Fig. 9C), in agreement with the finding that cadherin binding to β-catenin strengthens adhesion (Chu et al., 2004). Overall, there was a relatively weak correlation between ligand adhesiveness and axonal growth (Fig. 9D). This observation supports the notion that, beyond binding, N-cadherin coupling to the cytoskeleton is a clearer determinant of growth cone advance.
We addressed here the issue of how N-cadherin adhesion couples to actin dynamics to allow for the locomotion of neuronal growth cones. We specifically tested the clutch model, which views the retrograde actin flow as a motor that idles when it does not encounter adhesion to connect to, but which can engage into forward motion providing sufficiently strong transmembrane coupling to substrate-immobilized ligands (Mitchison and Kirschner, 1988; Suter and Forscher, 1998). We used primary neurons interacting with plane substrates or microspheres coated with purified recombinant N-cadherin. Such a biomimetic system allows a precise control of the type and density of ligand molecules presented to the cells and of the initial time of the interaction, which is not possible in natural contacts that form at arbitrary moments and where different types of adhesion proteins can coexist in unknown stoichiometry. The experimental design is summarized in Figure 10.
Our main result is a strong correlation between the rate of growth cone advance on N-cadherin and the mechanical coupling between ligand-bound N-cadherin receptors and the actin flow, in conditions of varying ligand density and expressing mutated N-cadherin receptors. In contrast, when N-cadherin binding alone was measured using bead adhesion assays, we observed a much weaker correlation between ligand adhesiveness and axonal growth, confirming previous results reported for DRG neurons plated on N-cadherin, L1, or polylysine substrates (Lemmon et al., 1992). Thus, our data support a direct mechanical linkage between ligand-bound N-cadherin receptors and the rearward-moving actin network, allowing traction forces to be transmitted to the substrate and resulting in forward progression (Mitchison and Kirschner, 1988; Suter and Forscher, 1998). The coupling process is likely to be regulated selectively for each adhesion molecule, as demonstrated for the IgCAM member L1, which can engage the actin flow through ankyrin as neurites are initiated, but not as axons extend (Gil et al., 2003; Nishimura et al., 2003). Such spatial control seems to be specific for L1, because L1 is also actively recycled from the base to the tip of growth cones through clathrin-dependent trafficking, and thereby exhibits a distribution gradient at the growth cone surface (Kamiguchi and Yoshihara, 2001; Dequidt et al., 2007).
Here, we focused on the later stages of axon elongation and showed that only N-cadherin had a strong ability to both promote growth cone translocation and couple to the actin flow. Indeed, the control molecule cadherin-11 showed very poor coupling but slightly promoted axonal growth (Fig. 2H), an effect that may involve a transduction pathway involving extracellular cis-binding to the FGF receptor (Saffell et al., 1997; Boscher and Mege, 2008). Conversely, SynCAM, which was reported to act as a trans-synaptic bridge through strong homophilic binding (Biederer et al., 2002; Breillat et al., 2007), coupled rather well to the actin flow but was completely unable to sustain growth cone progression. One possibility to explain this result is that synCAM adhesions are so robust that they cannot detach, and thereby impair growth cone advance. Such effect was reported for fibroblasts migrating on fibronectin, where at high levels of cell-substratum adhesiveness, cell speed is rate limited by the disruption of cytoskeleton–integrin–extracellular matrix linkages at the rear (Palecek et al., 1997). In that respect, the lifetime of N-cadherin adhesions at neuronal growth cones measured by fluorescence recovery after photobleaching (FRAP) experiments on Ncad-Fc-coated bead contacts lies in the range of several tens of minutes (Thoumine et al., 2006). This precisely fits the duration it would take for an adhesion formed at the growth cone periphery to detach at the rear (i.e., ∼10 μm further), the growth cone traveling at a velocity of 0.1–1 μm/min. Such relatively rapid turnover of N-cadherin adhesions seems to be especially adapted to allow the locomotion of growth cones from hippocampal neurons. As another example, the NcadAAA mutant, which shows a threefold higher homophilic bond lifetime than NcadWT (Thoumine et al., 2006), also shows a reduction in neurite progression that may be linked to high adhesiveness.
Transfection of N-cadherin receptors mutated in catenin-binding regions reduced both cytoskeletal coupling and neurite extension, demonstrating that the clutch directly involves catenin partners. The NcadΔextra mutant, which carries only the transmembrane domain and intracellular tail of N-cadherin, acts as a direct competitor at the cell surface for N-cadherin interactions with all intracellular partners, and as such had the most pronounced and specific effects in these assays. We attempted to use NcadΔβcat and NcadAAA constructs to more selectively perturb catenin interactions, reasoning that by competing for Ncad-Fc ligand binding on the beads, these mutated receptors should reduce the coupling between endogenous N-cadherin receptors and actin, and consequently growth cone velocity on Ncad-Fc. Indeed, β-catenin is well known to physically link cadherin receptors to actin filaments through binding to α-catenin (Mege et al., 2006). Conversely, p120 catenin is known to modulate cadherin adhesiveness (Ozawa and Kemler, 1998; Yap et al., 1998; Thoumine et al., 2006) and was reported to bind cortactin (El Sayegh et al., 2004), providing a potential parallel pathway to connect N-cadherin to the actin network. However, the effects of NcadΔβcat and NcadAAA were not completely specific of the Ncad-Fc substrate, as these mutants also reduced neurite extension on laminin-coated glass. This may be attributable in part to a potential cross talk between cadherins and integrins (e.g., the juxtamembrane domain of cadherin where p120 binds was shown to regulate integrin-mediated adhesion and neurite outgrowth) (Lilien et al., 1999), and N-cadherin is part of a protein complex including NCAM, which stimulates neurite outgrowth by an integrin- and FGF receptor-dependent mechanism (Cavallaro et al., 2001). Nevertheless, NcadΔβcat also exhibits a higher adhesive turnover rate as measured by FRAP (Thoumine et al., 2006), resulting in reduced adhesiveness (Fig. 9), and as such does not act as a good competitor for ligand binding. Furthermore, we noted a systematic reduction of Ncad-Fc bead velocity in growth cones of NcadAAA-transfected neurons, reflecting a slower actin flow rate, which may explain the higher inhibitory effect of NcadAAA on neurite length. One mechanism by which expression of NcadAAA could slow down the retrograde actin flow would be to leave a fraction of p120 free to interact with other partners, such as small GTPases (Rac/Rho), which are known to affect actin dynamics (Elia et al., 2006). Alternatively, because p120 is implicated in the recycling of N-cadherin at the plasma membrane (Chen et al., 2003; Davis et al., 2003), it is possible that the reduction of growth cone advance observed for the NcadAAA mutant on Ncad-Fc substrate is attributable to such altered trafficking. Overall, these observations may explain results from in vivo experiments on retinal ganglion cells, which showed that the N-cadherin intracellular tail region upstream of the β-catenin binding site, where p120 binds, was critical for axonogenesis (Riehl et al., 1996).
Thus, the use of mutated receptors reaching some limits, we adopted a different strategy using siRNA against α-catenin. The efficient downregulation of α-catenin in neurons led to a specific inhibition of neurite extension on Ncad-Fc substrates and coupling to actin, which represents the best evidence for catenin implication in the clutch mechanism. These data fit nicely with the newly emerging role for α-catenin as a key molecule at the interface between intercellular adhesion and actin dynamics (Kobielak and Fuchs, 2004; Drees et al., 2005).
By showing recapture events in the optical trap, we also provide strong evidence for the continuous slippage of bonds between N-cadherin receptors and actin at low forces. Since those beads coated with a small number of N-cadherin ligands are likely to attach to very few receptors (Thoumine and Meister, 2000), the retrapping events were interpreted as the breaking of a small number of molecular bonds. Because the forces involved are much lower than the 30–100 pN interaction between individual cadherin molecules (Perret et al., 2004; Pittet et al., 2008), and because beads remain attached to the growth cone surface when the trap is stopped, rupture is unlikely to occur between N-cadherin ligands and receptors. Instead, the 1 pN retrapping events more likely correspond to the rupture of single bonds between the cadherin–catenin complex and actin, which were proposed previously to be highly dynamic (Drees et al., 2005). The higher force values may represent multiples of the unitary 1 pN value, and correspond to the simultaneous breakage of several bonds. The forces involved here are similar to the 2 pN bond between integrin trimers and the actin cytoskeleton through talin reported in fibroblasts (Jiang et al., 2003), suggesting that slippage at low forces is a general mechanism underlying actin anchorage to substrate-bound ligands.
By applying higher forces, we were able to restrain microspheres and locally stop the actin flow. We then observed striking accumulation of actin-GFP at the bead contact, which was also seen at sites of N-cadherin clusters touching the substrate. The fact that in both assays, actin accumulated to greater levels than N-cadherin receptors confirms the different dynamics of actin and cadherin-catenin complexes at adhesion sites (Yamada et al., 2005) and suggests some sort of amplification, not only a passive recruitment of pre-existing actin filaments. We propose a mechanism based on a local actin polymerization driven by mechanical force applied at N-cadherin receptor clusters, as observed in myogenic cells using patterned force detection with micropillar arrays (Ganz et al., 2006). It is possible that nascent N-cadherin adhesions may initiate such process by mobilizing α-catenin molecules otherwise in competition with the Arp2/3 nucleating complex (Drees et al., 2005). Indeed, an accumulation of α-catenin-GFP was also observed in the microneedle assay (data not shown) and the recruitment of actin at microspheres was prevented by overexpression of the NcadΔextra mutant, which competed for the binding of catenin partners. Alternatively, actin accumulation may be linked to other pathways, for example, activation of Rac1, which we previously reported to mediate N-cadherin anchoring to the actin flow (Lambert et al., 2002), or FGF receptor activation and downstream signaling through the phospholipase C-γ/DAG lipase pathway and calcium regulation (Saffell et al., 1997; Boscher and Mege, 2008). In any case, local actin accumulation at stiff N-cadherin contacts may serve to strengthen the adhesive complex and act as a driving force for growth cone extension downstream of N-cadherin liganding.
In conclusion, our data suggest that N-cadherin exerts a neuritogenic effect in neurons from the CNS of vertebrates by stimulating growth cone migration via a molecular clutch mechanism, catenins being part of the force-transducing complex to actin filaments.
This work was supported by the French Ministry of Research, Centre National de la Recherche Scientifique, Conseil Régional Aquitaine, Association Franc̨aise contre les Myopathies, Association pour la Recherche sur le Cancer, and Inserm. We thank C. Dequidt for preliminary experiments, C. Breillat for the production of SynCAM-Fc, P. Gonzales for coverslip preparation, D. Bouchet and B. Tessier for neuronal cultures, V. Marthiens for the production of cad11-Fc, and J. Falk, G. Giannone, D. Perrais, and C. Sarrailh for critical reading of this manuscript.
- Correspondence should be addressed to Olivier Thoumine at the above address.