INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

The nerve growth cone is the sensory motile organelle that translates extracellular cues into axon guidance via directed motility (reviewed by Tanaka and Sabry, 1995; Jay, 1996). They do so by localized changes in motility at the leading edge that bias the movement of the rest of the growth cone (Jay, 1996). It is clear from studies in vitro (Marsh and Letourneau, 1984) and in vivo (Bentley and Toroian-Raymond, 1986) that F-actin is critical for directing growth cone motility. Specific actin-associated proteins are required to control F-actin at the leading edge, but it has been difficult to show which of the many actin-associated proteins act at that location (Letourneau, 1996).

Among the best candidates to act at the leading edge of the growth cone is radixin (Gonzalez-Agosti and Solomon, 1996). Radixin is a prototypic member of the ERM (ezrin-radixin-moesin) family of proteins (reviewed by Arpin et al., 1994; Tsukita et al., 1997). It was initially identified as a barbed end capping protein (Tsukita et al., 1989) and was later shown to be highly homologous to ezrin and moesin (Funayama et al., 1991; Sato et al., 1991). Cryptic F-actin-binding sites within the carboxyl-terminal domain were revealed by denaturation studies and by expression of single domains of radixin in vitro (Henry et al., 1995; Magendantz et al., 1995). Similar sites have also been seen with ezrin and moesin (Gary and Bretscher, 1993; Turunen et al., 1994; Martin et al., 1995). Radixin and other ERM proteins also bind to membrane proteins such as CD44 that, in turn, can bind to hyaluronic acid in the extracellular matrix (Tsukita et al., 1994). From these studies, it has been hypothesized that the ERM proteins work as plasma membrane-actin filament cross-linkers (Algrain et al., 1993; Tsukita et al., 1994). Radixin is the predominant ERM family member in chick sympathetic neuronal growth cones, and its localization is diminished after induction of growth cone collapse (Gonzalez-Agosti and Solomon, 1996). Together, these studies suggest that radixin has a role in growth cone shape and motility but, thus far, direct evidence to support this hypothesis has been lacking.

It has been difficult to show that any specific ERM protein functions in cells. ERM proteins are highly homologous with each other and are often coexpressed (Sato et al., 1991). They colocalize and bind to similar sites (Henry et al., 1995) and may have overlapping functions (Takeuchi et al., 1994). Overexpression (Henry et al., 1995) and misexpression (Martin et al., 1995) of radixin or ezrin domains cause cell shape changes, but these studies have not shown specific roles for any one of the ERM proteins. Reduction of all three ERM proteins by antisense oligonucleotide expression showed defects in cell attachment of mouse epithelial cells, but antisense expression of any one ERM protein had little effect (Takeuchi et al., 1994). These experiments did, however, suggest a more specific role for moesin in microvilli formation in thymoma cells (Takeuchi et al., 1994).

Recently, essential roles for ezrin were shown using microscale chromophore-assisted laser inactivation (micro-CALI) to inactivate ezrin in fibroblasts (Lamb et al., 1997). Micro-CALI inactivates a protein within a 10-µm region by targeting a 620-nm laser microbeam via a specific antibody that is labeled with the dye Malachite green (MG) (Jay, 1988; Diamond et al., 1993). The acute and localized inactivation of ezrin resulted in a loss of membrane ruffling and pseudopodial retraction in transformed fibroblasts, and in marked retraction of the leading edge of normal fibroblasts. This study suggests that ezrin has critical roles in fibroblast cell shape and motility (Lamb et al., 1997). It also suggests that a similar strategy would be useful to address the role of radixin in growth cone shape and motility. In the present study, we applied micro-CALI to show a role for radixin in embryonic chick dorsal root ganglion (DRG) neuronal growth cones.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

Chick DRG Neuronal Culture

Dissociated chick DRG neurons were prepared as described by Sydor et al. (1996). DRGs were dissected from E10-12 chick embryos and suspended in 0.25% trypsin in HBSS for 8 min. After trypsinization, cells were spun down at 1000 × g for 2 min, and the cell pellet was resuspended in chick neuronal media (L-15 media from Sigma Chemical, St. Louis, MO, supplemented with 10% fetal bovine serum from Hyclone, Logan UT). Cells were triturated using an up-and-down motion of a 200-µl Gilson Pipetteman, set at 40 µl (~80 gentle strokes). They were then plated on donut dishes (30-mm culture dishes with 11/16 in. drill holes affixed with coverslips coated with poly-L-lysine and laminin) with 3 ml of chick neuronal media. Cells were incubated at 37°C in an air incubator for 1.5-2 h before experimentation.

Antibodies

Affinity-purified anti-radixin antibodies were generous gifts of Frank Solomon and Etchell Cordero (Massachusetts Institute of Technology, Cambridge, MA) and have been previously described by Winckler et al. (1994). Rabbit polyclonal antibody 457-3 is specific for radixin alone and was raised against amino acids 400-409 in the carboxyl-terminal domain of radixin. Rabbit polyclonal antibody 220, which recognizes ezrin, radixin, and moesin, was raised against the first 13 amino acids at the amino terminus that are common to all three ERM proteins. We used the following secondary antibodies (from Cappel Laboratories, Malvern PA): rhodamine- or fluorescein-conjugated goat anti-rabbit IgG; and fluorescein-conjugated rabbit anti-mouse IgG. For micro-CALI experiments, antibodies were labeled with MG as previously described by Beermann and Jay (1994). Labeling ratios of MG-457-3, MG-220, and MG-IgG were ~6-8 dye moieties/IgG molecule.

Antibody Loading

For both micro-CALI and immunocytochemistry experiments, neurons were loaded via trituration with MG-labeled or unlabeled primary antibodies as described by Sydor et al. (1996). After trypsinization, antibody solutions (50 µl) at 2 mg/ml in HBSS with 1 mg/ml fluorescein-labeled IgG (as a tracer) were added to six whole DRGs during trituration. After trituration, cells were separated by 2 min of centrifugation at 1000 × g. The antibody solution was removed and cells were gently resuspended in chick neuronal media before plating. Trituration creates temporary holes in the cell membrane, enabling antibodies to enter (Borasio et al., 1989) and has been used to load antibodies for micro-CALI previously (Chang et al., 1995; Sydor et al., 1996).

Immunocytochemistry

Cultured chick DRG neurons on glass coverslips were fixed with 3.7% paraformaldehyde in PBS for 25 min at 37°C, washed three times with PBS, permeabilized by 0.05% Triton X-100/PBS for 15 min at room temperature, and blocked with 10% fetal calf serum in PBS for 30 min. Fixed neurons were incubated overnight in primary antibodies (1:500 dilution of 457-3; 1:1000 for 220), washed extensively, and then probed with rhodamine- or FITC-conjugated secondary antibodies (1:100, diluted in blocking buffer) for 2 h at room temperature. After five 10-min rinses in PBS, coverslips were mounted in Fluoromount-G (Southern Biotechnology Associates) and were then viewed with a Zeiss confocal microscope (model LSM 410; Carl Zeiss, Thornwood, NY).

Immunoblotting

Immunoblot analysis was performed according to Dubreil et al. (1987). In brief, chick DRGs were collected and placed in 200 µl of HBSS (without calcium or magnesium) and spun down briefly at 1000 × g. HBSS was removed, and 2% boiling SDS sample buffer was added. The sample was then boiled for 3 min, sonicated at 70% duty cycle for 3 min, and boiled for an additional 5 min. DRG lysates (4 DRGs per lane) were separated by electrophoresis on 7.5% SDS-PAGE minigels and transferred by electrophoresis to nitrocellulose at 40 V for 1.5 h. Nitrocellulose blots were then blocked with 5% nonfat dry milk in 0.03% Triton X-100 in PBS for 1.5 h at room temperature. They were incubated for 2.5 h at room temperature with 457-3 (1:500) and 220 (1:1000) and probed with alkaline phosphatase-labeled goat anti-rabbit IgG antibodies (1:500) for 1 h at room temperature. Blots were then developed in alkaline phosphatase substrate (Sigma Chemical, St. Louis MO).

Micro-CALI of Proteins in Growth Cones

Micro-CALI was performed on chick DRG growth cones 1-6 h after plating according to Wang et al. (1996). Throughout the micro-CALI experiments, DRG cultures were maintained at 37°C with a stage incubator (Opti-Quip, Highland Mills, NY). In a typical micro-CALI experiment, antibody-loaded neurons were selected by epifluorescence. A selected growth cone was observed by video-enhanced time-lapse microscopy (Scion Image software) for 5 min (recorded every 15 s) as described by Wang et al. (1996). A region of the growth cone was laser irradiated for 5 min using a nitrogen-driven pulsed dye laser (model 337, Laser Science, Newton MA) with the fluorescent laser dye DCM and observed for an additional 10 min.

Methods of Quantitation

Results of asymmetrical (half-growth cone) micro-CALI studies were measured using NIH Scion System Software. Lamellipodial area was measured in every frame of the time-lapse period and plotted as a function of time. The rates of retraction were obtained by taking the first derivative of these plots. Area retraction was defined as a percent decrease in the lamellipodial area (µm²) in laser-irradiated or unirradiated regions during 5 min of observation. The rates of lamellipodial retraction/extension and the percent change in lamellipodial surface area inside and outside the irradiated region were compared 5 min before irradiation to ensure that they were statistically indistinguishable. Additionally, neurite length was measured before, during, and after laser irradiation using NIH Scion Imaging software. A fixed point on the base of the cell body was chosen as a landmark at t = 0 min. All subsequent measurements began at this fixed point and ended at the base of the growth cone body. Neurite extension rate was defined as the change in neurite length (µm/min) during the period of time-lapse observation. Results of symmetrical growth cone studies were scored as splitting versus not splitting after micro-CALI. A split growth cone was defined as a growth cone whose laser-irradiated region retracted but whose unirradiated regions continued to grow and diverge into two smaller growth cones.

All quantitative data are reported as mean ± SD. These data were obtained from more than 10 experiments and a total of more than 20 growth cones for each experimental condition. Statistical analysis was performed with Stat View II (Abacus Concept, Berkeley, CA). Analysis of the significance between percent change of lamellipodial area was assessed using Student's two-tailed, paired t test for asymmetric application of micro-CALI, and unpaired t test for comparison between samples using MG-labeled nonimmune IgG- and Mg-labeled anti-radixin-loaded growth cones for micro-CALI. Analysis of variance and Poisson test were also used to assess significance between multiple sets of data.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

The hypothesis that radixin has a critical role in growth cone motility (Gonzalez-Agosti and Solomon, 1996) was tested in DRG neuronal growth cones in culture using micro-CALI. This approach allowed us to observe changes in growth cone morphology and motility in response to the acute and localized loss of radixin. To perform micro-CALI, a specific antibody is tagged with MG dye, which is then introduced into DRG neurons by trituration, and the neurons are allowed to extend neurites in culture. An area of inactivation within the growth cone is selected, and irradiated with a laser microbeam at a wavelength of 620 nm. The MG label absorbs this laser light to generate short-lived free radicals (Liao et al., 1994) that selectively inactivate proteins bound by the MG-labeled antibody. CALI takes advantage of the specificity inherent in tight binding interactions, as unbound MG-labeled reagents do not cause significant damage (reviewed by Wang and Jay, 1996).

Specificity of Antibodies

For these studies we employed two polyclonal antibodies: 457-3, which recognizes the carboxyl-terminal domain of radixin alone (Winckler et al., 1994); and 220, which recognizes the amino terminus of all three ERM proteins. We tested their specificity by immunoblot analysis and immunocytochemistry. Immunoblotting confirmed that 457-3 and 220 specifically recognized ERM species in DRG neuronal lysates (Figure 1A). Immunoblotting of DRG cell lysates with 457-3 detected a predominant 82 kDa species, which corresponds to the molecular mass of radixin. Immunoblotting of DRG cell lysates with 220 showed two bands of similar intensity at 82 kDa and 75 kDa, which correspond to the apparent molecular weights of radixin and moesin, respectively. It is possible that the moesin comes from contaminating fibroblasts in the lysate, even though DRGs have many more neurons than fibroblasts. This is unlikely because a band corresponding to ezrin (85 kDa), which is abundant in fibroblasts (Lamb et al., 1997), was not detected in these immunoblots. The lack of this band shows that fibroblast contamination was small and suggests that moesin is present in DRG neurons.

View larger version (65K): [in this window] [in a new window]   Figure 1.   Specificity of anti-radixin antibodies. (A) DRG neuronal lysates analyzed by immunoblot with rabbit polyclonal antibodies 220 (lane a) and 457-3 (lane b). Antibody 457-3 (which is specific for radixin) recognizes a single band corresponding to radixin. Antibody 220 (which binds to all three ERM proteins) recognizes two bands with molecular weights corresponding to radixin and moesin. DRG neuronal lysates do not appear to express ezrin. (B) DRG-dissociated neurons were grown on laminin-poly-L-lysine-coated coverslips and were fixed and probed with 457-3 anti-radixin followed by FITC-labeled anti-rabbit IgG. Radixin is present throughout the growth cone and exhibits a punctate pattern. Staining was undetectable for controls that lack 457-3.

Immunocytochemistry with 457-3 was used previously to show the presence of radixin in the chick sympathetic neurons (Gonzalez-Agosti and Solomon, 1996). We used this assay to show that radixin is also found in the growth cones of chick DRG neurons (Figure 1B). Growth cones were brightly stained, as were cell bodies and neurites. Within the growth cone, radixin staining was uniform and exhibited a punctate pattern (Figure 1B) that extended out from the lamellipodia into the filopodia. This pattern was similar to the staining observed with two other anti-ERM antibodies previously reported: 904, which recognizes all ERM isoforms (Birgbauer et al., 1991); and 3D11, which recognizes radixin (Everett and Nichol, 1990). Immunocytochemistry with 220 showed similar but heavier staining throughout the growth cone, cell body, and neurites (data not shown) most likely reflecting the ability of this antibody to recognize both radixin and moesin in DRG neurons. The pattern observed was different than that seen using 13H9 with DRG neurons; this antibody brightly stained filopodia (Goslin et al., 1989; Birgbauer et al., 1991). While 13H9 detects an ezrin-like immunoreactivity, it does not recognize ERM proteins by immunoblot (Gonzalez-Agosti, personal communication).

Together, these experiments show that 457-3 and 220 have the specificity required for their use for micro-CALI of radixin in DRG growth cones. As no other bands are detected by immunoblotting using 457-3, this antibody is unlikely to bind to other proteins in DRG neurons. Thus, the effects of micro-CALI using 457-3 would be specific for radixin. Micro-CALI using 220 would be expected to inactivate both radixin and moesin in DRG neurons. The use of 457-3 and 220 for micro-CALI also allows us to target two different domains of radixin.

Micro-CALI of Radixin Causes Lamellipodial Retraction

Micro-CALI of radixin using MG 457-3 resulted in a rapid lamellipodial retraction of the irradiated half. Figure 2 shows an example of this. There was a marked retraction of lamellipodia in the irradiated half during the laser period, t = 0-5 min, but no retraction of the unirradiated half (Figure 2, B-D). Filopodia did not appear to be affected. They maintained their structure throughout this period of lamellipodial retraction. The growth cone pictured in Figure 2 showed a turning away from the irradiated region but this is not always so. The asymmetric loss of radixin may direct subsequent neurite outgrowth, but recovery is rapid (see below), and a short period of loss of function may not be sufficient to cause turning. Indeed, the localized inactivation of myosin I must be performed several times to evoke turning (Wang and Jay, unpublished data).

View larger version (33K): [in this window] [in a new window]   Figure 2.   Micro-CALI of radixin causes lamellipodial retraction. A chick DRG neuron was loaded with MG-labeled 457-3 and plated on a laminin-coated coverslip to allow a growth cone to emerge and time-lapse imaging was begun at t = 5 min (A). The growth conewas subjected to laser irradiation on one side of the growth cone (white circle) beginning at t = 0 min (B) and ending at t = 5 min (C), and the growth cone was observed for another 10 min (D and E). The lamellipodia in the irradiated half retracts (B-D) and begins to recover by t = 10 min (E). Scale bar, 10 µm.

In contrast, laser irradiation of growth cones loaded with dye-labeled nonimmune IgG (MG IgG) had no effect on lamellipodial area (Figure 3). Normal growth continued and normal growth cone morphology was maintained throughout the growth cone (Table 1). There was no observable difference between irradiated and nonirradiated sides of the growth cone. Loading of MG-labeled anti-radixin antibodies also did not affect lamellipodia by themselves. The changes of lamellipodial area for growth cones loaded with MG-anti-radixin or MG-nonimmune IgG were indistinguishable without laser irradiation (Table 1, Prelaser and Postlaser). This shows that the MG-457-3 by itself did not affect function with respect to growth cone motility (although other roles of radixin may have been affected).

View larger version (34K): [in this window] [in a new window]   Figure 3.   Micro-CALI using MG-nonimmune IgG has no effect. A chick DRG neuron was loaded with MG-labeled nonimmune IgG and plated on laminin-coated coverslip to allow a growth cone to emerge. Time-lapse imaging was begun at t = 5 min (A). The growth cone was subjected to laser irradiation within a small region of the growth cone (white circle) beginning at t = 0 min (B) andending at t = 5 min (C), and the growth cone was observed for another 10 min (D and E). The lamellipodia within the irradiated region do not retract (B-E). Scale bar, 10 µm.qending at t = 5 min (C), and the growth cone was observed for another 10 min (D and E). The lamellipodia within the irradiated region do not retract (B-E). Scale bar, 10 µm.

Table 1 shows the quantitation of lamellipodial retraction caused by micro-CALI of radixin or by irradiation of MG-labeled nonimmune-IgG-loaded neurons. In general, lamellipodia extend and retract over time such that there is no net average change in lamellipodial area. This was observed for all control treatments, and these values range from ~1.5 to 2% of the initial lamellipodial area. These values are statistically indistinguishable from each other (p > 0.2). In contrast, regions of DRG growth cones treated with micro-CALI of radixin showed a ~30% decrease in lamellipodial area within the laser spot. This decrease was significantly different than lamellipodial change on the unirradiated sides of growth cones (n = 24, p < 0.0001 by paired t test) and was also significantly different from all control treatments (p < 0.0001 by analysis of variance).

These experiments showed that binding of the MG-457-3 to radixin together with laser irradiation was responsible for the lamellipodial retraction observed. Irradiation of growth cones loaded with MG-nonimmune antibodies also had no effect on lamellipodia. Importantly, there was no significant difference between the irradiated and nonirradiated regions of micro-CALI-treated growth cones before or after laser irradiation. That there was no significant difference before laser irradiation shows that there was no bias in selecting regions to be irradiated. That there was no significant difference after laser irradiation shows that growth cones recover rapidly, likely due to the movement of active radixin from unirradiated regions. Together these data show that micro-CALI of radixin causes lamellipodial retraction in neuronal growth cones.

Micro-CALI of Radixin Causes Growth Cone Splitting

When micro-CALI of radixin was directed to the middle of the leading edge, growth cones often split (Figure 4). The leading edge retracted while the unirradiated regions grew and extended normally. Continued motility divided the growth cone into two separate growth cones (Figure 4, D and E) each with a nascent neurite. The time course of growth cone splitting after micro-CALI was rapid with a t_1/2 of ~2 min (Figure 5). Although growth cone splitting can occur spontaneously in vivo (Letourneau et al., 1986), in these experiments, it occurred precisely during laser irradiation, suggesting that it was a direct result of micro-CALI of radixin. Table 2 shows the percentage of growth cones that split for every experimental and control condition. More than 80% of the growth cones showed splitting during irradiation of MG 457-3-loaded neurons. This splitting frequency was significantly different than that observed for all control treatments (p < 0.0001 by Poisson test). In contrast, only one growth cone during laser irradiation in a population of cells loaded with MG IgG (4.5%; p > 0.5 by Poisson test). Splitting was also strictly dependent on the dye-labeled antibody. Untreated cells, cells loaded with unlabeled 457-3 and 220, or cells loaded with MG-nonimmune IgG showed no splitting regardless of laser light.

View larger version (39K): [in this window] [in a new window]   Figure 4.   Micro-CALI of radixin in the central region causes growth cone splitting. A chick DRG neuron was loaded with MG-labeled 457-3 and plated on a laminin-coated coverslip to allow agrowth cone to emerge. Time-lapse imaging was begun at t = 5 min (A). The growth cone was subjected to laser irradiation in its central region (white circle) beginning at t = 0 min (B) and ending at t = 5 min (C), and the growth cone was observed for another 10 min (D and E). The lamellipodia in the irradiated region retract (B and C), and the growth cone splits into two small growth cones (D and E). Similar results were seen using MG-labeled 220 in place of MG-labeled 457-3 (our unpublished results). No effect was seen with MG-labeled nonimmune IgG (see Table 2). Scale bar, 10 µm.

View larger version (16K): [in this window] [in a new window]   Figure 5.   The effects of micro-CALI of radixin and of all ERMs are indistinguishable. The time course for growth cone splitting was plotted for all samples subjected to micro-CALI of radixin alone and micro-CALI of all ERMs. Growth cone splitting is initiated early during laser irradiation and occurs rapidly for both treatments.

Micro-CALI of Radixin Using Two Different Antibodies Has Similar Effects

Thus far, micro-CALI has been directed to radixin via 457-3, which recognizes the carboxyl-terminal domain of radixin specifically (Winckler et al., 1994). To further test our hypothesis of radixin function, we applied micro-CALI using MG-labeled 220, which binds to radixin at its amino terminus. The effects of micro-CALI using MG-labeled 220 were indistinguishable from those observed after micro-CALI using MG-labeled 457-3 (Table 2). The time course of growth cone splitting is also indistinguishable for these two samples (Figure 5). No additional changes in growth cone behavior were seen using MG-labeled 220 for micro-CALI. Thus, micro-CALI, using two different antibodies that bind to different domains of the radixin polypeptide, causes similar effects on growth cone behavior. These findings suggest that the effects of micro-CALI are specific and support the hypothesis that radixin is involved in lamellipodial stability. However, it is also possible that the amino- and carboxyl-termini are close together in radixin's three-dimensional structure, and it has been suggested that they associate in native radixin (Magendantz et al., 1995).

The 220 antibody also binds to ezrin and moesin, and moesin is also present in DRG neurons (Figure 1A). As such, if moesin had a different role in growth cone motility, we might have expected to see different growth cone behaviors in response to micro-CALI using MG-labeled 220. We did not see this. This suggests that disruption of radixin is sufficient to cause growth cone splitting and that moesin may have no additional role in lamellipodial stability. It is formally possible that CALI using MG-labeled 220 does not affect moesin, but given the similar structures of moesin and radixin, we view this as unlikely.

Micro-CALI of Radixin Does Not Affect Neurite Extension

An additional demonstration of the specificity of micro-CALI of radixin was provided by comparing neurite extension rates of neurons, loaded with MG-labeled 457-3, 220, and IgG, before, during, and after asymmetric laser irradiation. We also examined the extension rates of neurons loaded with unlabeled 457-3 and 220 (Table 3). Neurite extension rates for all of these treatments (regardless of laser light) were not significantly different than the rate of neurite extension for untreated neurons (p > 0.15 for all pairwise comparisons during or after irradiation). These data show that the acute loss of radixin by micro-CALI does not affect neurite extension. They provide evidence that this treatment does not cause nonspecific damage to growth cones that results in their collapse.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

We have shown that the acute loss of radixin generated by micro-CALI caused lamellipodial retraction within the region of inactivation and growth cone splitting when micro-CALI was targeted to the middle of the leading edge. Growth cones are held under tension as they move (Bray, 1979), and this tension is important for subsequent neurite outgrowth (Bray, 1984). Letourneau et al. (1986) proposed that a critical step in branching is the spreading of the neurite cytoskeleton by tension generated at the lateral margins of the leading edge of the growth cone. Our findings are consistent with radixin maintaining the stability of lamellipodia that resists this tension. When radixin's function is diminished at the leading edge, this may cause a local weakening of the cytoskeleton in the irradiated region. Subsequent movement splits the growth cone at this site to generate branching into two new growth cones. As such, a localized change in radixin function at the leading edge can affect the pattern of neurite outgrowth.

The Effects of Micro-CALI of Radixin Are Specific

The effects of micro-CALI of radixin on lamellipodia were not likely due to nonspecific damage of the cytoskeleton. Micro-CALI of radixin using antibodies that recognize different domains of radixin had similar effects. CALI-induced damage is spatially restricted with a half-maximal radius of damage of 15 Å around each dye moeity (Liao et al., 1994). This approach has been used to inactivate single subunits in the T cell receptor with only slight effects on nearest neighbors in the complex (Liao et al., 1995). CALI is spatially restricted even between different domains of a single protein. For example, CALI using an antibody that recognizes the tail and neck domain of myosin V inhibits in vitro motility associated with the neck domain without affecting the actin-dependent ATPase activity associated with the head domain (Wang et al., 1996). CALI has mimicked genetic loss of function in Drosophila in all cases tested directly (Schmucker et al., 1994; Schroeder et al., 1996). Micro-CALI performed on a variety of actin-associated proteins in nerve growth cones has shown distinct effects on motility for each protein studied (Sydor et al., 1996; Wang et al., 1996). For example, micro-CALI of myosin I caused lamellipodial expansion into the laser spot (Wang et al., 1996). Finally, micro-CALI targeted directly to actin in growth cones using MG-labeled G-actin had no effect on motility (Sydor, 1995). Although it not possible to entirely rule out that CALI damages neighboring proteins for any specific case, these studies suggest that this is unlikely to occur.

Radixin and Ezrin May Play Similar Roles in Different Cell Types

A recent study using micro-CALI of ezrin (Lamb et al., 1997) showed a role for ezrin at the leading edge of transformed fibroblasts. This caused pseudopodial retraction that was similar to the lamellipodial retraction that we observed during micro-CALI of radixin.

This suggests that radixin and ezrin play similar roles in different cell types to stabilize the leading edge of moving cells. However, when ezrin was inactivated in normal fibroblasts, there was a marked and rapid retraction of the entire leading edge when micro-CALI was targeted here. This retraction left fibers remaining on the substrate like those seen on the trailing edge of moving fibroblasts or when these cells detach (Harris, 1994). This difference may be due to the fact that the traction forces that hold cells to the substratum are much stronger for normal fibroblasts (2 × 10⁷ N; Harris et al., 1980; Oliver et al., 1995) than for growth cones (< 6 × 10⁹ N; Lamoureux et al., 1989).

It has been suggested that radixin and ezrin have redundant function and are coexpressed in many cell types as a safety measure (Tsukita et al., 1997). Our studies support this hypothesis. This may explain why it was necessary to suppress all three ERM proteins in cultured epithelial cells to affect cell attachment (Takeuchi et al., 1994). Chick DRG neurons do not express ezrin (Figure 1A), and thus the effects of the loss of radixin may be more pronounced in neurons compared with other cells. Alternatively, the difference in severity observed here compared with antisense experiments may be due to the nature of disruption of the two methods. Antisense RNA generates the chronic and global loss of cellular expression of any particular protein. It may allow compensation of proteins with overlapping function although no increase in expression was observed in these antisense experiments (Takeuchi et al., 1994). Micro-CALI allowed us to observe localized changes in motility and growth cone shape during inactivation. This analysis likely precludes compensation. Thus, micro-CALI provides a good test of a protein's function during motility.

Radixin's Role in Lamellipodia

Our findings show that radixin is involved in lamellipodial stability but do not demonstrate how radixin acts here. Cells must regulate the connections of the cytoskeletal network to effect cell shape changes during motility (Ingber et al., 1994). These connections may be maintained to resist tensile force or may give way to change cell shape (Mooney et al., 1995). Radixin is a good candidate to play this role in nerve growth cones because it is found in growth cones (Gonzalez-Agosti and Solomon, 1996) and has in vitro binding sites for F-actin (Algrain et al., 1993; Henry et al., 1995) and membrane proteins (Tsukita et al., 1994). It has been suggested that radixin may cross-link actin filaments or link F-actin to the plasma membrane (Algrain et al., 1993; Tsukita et al., 1994, 1997), and our findings are consistent with this notion. ERM binding to F-actin and membranes may be regulated by signal transduction (Bretscher, 1989; Nakamura et al., 1995; Hirao et al., 1996; MacKay et al. 1997; Potter et al., 1998). As such, radixin may act to translate signals from extracellular cues to change shape and motility at the growth cone's leading edge.

Radixin may also function in actin assembly by its barbed end capping activity (Tsukita et al., 1989). Our observations would be consistent with a loss of actin assembly at the leading edge causing localized lamellipodial retraction. Micro-CALI of radixin could modify radixin such that F-actin at the leading edge could not be uncapped. A loss in adhesion at the leading edge could also cause lamellipodial retraction, and it has also been suggested that ERM proteins may have a role in cell adhesion (Takeuchi et al., 1994; MacKay et al., 1997). Establishing which of these cellular processes are critical for radixin's action in lamellipodia will require a detailed analysis of ultrastuctural changes in F-actin in response to micro-CALI of radixin.