RhoA and Rac1 GTPases mediate the dynamic rearrangement of actin in peripheral glia

Development of peripheral nervous system (PNS) glial cells in both vertebrates and insects is a dynamic process. Peripheral glia arise from the neural crest in vertebrates and the lateral edge of the central nervous system(CNS) in the embryo. During development, they migrate out over large distances to their peripheral axon targets (Carpenter and Hollyday, 1992; Sepp et al., 2000). After the cell migration phase has taken place, the peripheral nerves become ensheathed by glia. There has been intensive study of how glial cells wrap the peripheral nerves, especially in the vertebrate systems (Jessen and Mirsky,1999). By contrast, very little is known about how PNS glia carry out their migration phases.

The Drosophila embryo is an ideal model for the analysis of glial cell migration and nerve ensheathement. Mutant phenotypes of genes important in Drosophila peripheral glial migration can be readily detected,because the pattern of the peripheral nerves is very simple. In addition to their extensive migration, peripheral glia project long cytoplasmic sheaths around the axon tracts. Loss or disruption of the peripheral glial sheath results in defasciculated axon tracts, that is also an easily detectable phenotype (Sepp et al., 2001). As development of the peripheral glia in the Drosophila embryo has only recently been characterized, our knowledge of genes important in peripheral glial development is very limited. However, it is known that peripheral glia have an important role in mediating axon guidance during PNS development (Sepp et al.,2001).

As little is known about how peripheral glial cell migration occurs,studies on other cell types are useful to develop hypotheses on glial migration. Dynamic rearrangement of the actin cytoskeleton is crucial for the migration of multiple cell types (Hall and Nobes, 2000). In addition, interaction between the migrating cell and its underlying substrate is an important factor in a the ability of a cell to migrate (O'Connor and Bentley,1993; Lin and Forscher,1993). Thus, it is likely that actin rearrangements are involved in peripheral glial cell migration. As well, it is possible that neurons express molecules that stimulate the migration of glia along their axons by directly affecting actin dynamics.

The small Rho GTPases Rho, Rac and Cdc42 are mediators of actin dynamics in motile cells and during cell morphogenesis(Hall and Nobes, 2000). Each GTPase has been implicated in the formation of different Actin-based structures. In in vitro analysis of fibroblasts, Cdc42 is involved in extension of filopodia, Rho mediates stress fiber formation, and Rac is involved in membrane ruffling and the formation of lamellipodia(Nobes and Hall, 1995). The small Rho GTPases also have distinct functions in Drosophila neurons. Rac1 is involved in axonal outgrowth and steering, Cdc42 is involved in neuron morphogenesis and cell positioning, and RhoA is involved in neuroblast proliferation and is essential for dendritic but not axonal growth(Luo et al., 1994;Kaufmann et al., 1998;Lee et al., 2000;Hakeda-Suzuki et al., 2002;Ng et al., 2002). The current knowledge of the function of Rho GTPases in glia is limited to in vitro studies. Cdc42 is found to control cell polarity in astrocyte monolayers and affects cell motility in Schwann cell cultures(Cheng et al., 2000;Etienne-Manneville and Hall,2001). Rac1 is also involved in cell motility(Cheng et al., 2000), while RhoA activity affects cell morphology in cultured Schwann cells(Brancolini et al., 1999).

As Rho GTPases are widely expressed during embryogenesis, mutations in these genes affect many different tissue types. Therefore, in determining the effects of the GTPases in a specific tissue, standard mutant analysis is not always appropriate. We have chosen to analyze the function of Rho GTPases in peripheral glia during embryonic nervous system development ofDrosophila. For these experiments, we ectopically expressed mutant and wild-type constructs specifically in the glia using the GAL4/UAS gene expression system (Brand and Perrimon,1993). Distribution of actin-GFP was characterized in the wild-type and mutant backgrounds. The studies have shown that Rho and Rac have distinct activities in peripheral glial cell migration and nerve ensheathement, while Cdc42 did not appear to have a major role in peripheral glial cell development.

To image wild-type microtubule distribution, the rL82#29:GAL4driver (Sepp et al., 1999) was crossed to the UAS-tau-lacZ marker(Hidalgo et al., 1995). For imaging wild-type embryonic Actin distributions, a stock carrying therepo:GAL4 driver (Sepp et al.,2001) plus the UAS-actin-GFP marker(Verkhusha et al., 1999) was generated. To image microtubules along with Actin-GPF, therepo:actin-GFP line was crossed to the UAS-tau-lacZmarker.

For ectopic Rho GTPase expression, the repo::actin-GFP line was crossed to each of the following lines: UAS-RhoAV14(Fanto et al., 2000),UAS-RhoA (wild type) (Harden et al., 1999), UAS-RhoAN19(Strutt et al., 1997),UAS-DRac1wt, UAS-DRac1V12, UAS-Drac1N17, UAS-Drac1L89, UAS-Dcdc42V12and UAS-Dcdc42N17 (Luo et al.,1994). For constructs where multiple P-element transformants were available, the repo:actin-GFP line was crossed to the alternate transformant lines and phenotypes of the progeny were analyzed to verify that mutant phenotypes were due to ectopic expression of the UAS constructs rather than their genomic insertion sites.

For analysis of hypomorphic Rac and RhoA alleles, embryos were collected from the following stocks: RhoA^E3.10/CyO(Halsell et al., 2000),RhoAk02107a/CyO (Magie et al.,1999), Rac1J10/TM6 B, Tb, Rac1^J11/TM6 B, Tb,Rac2^Δ and Mtl^Δ/TM3, Sb(Hakeda-Suzuki et al., 2002;Ng et al., 2002). All embryos and larvae were raised at 21°C.

Embryos and larvae were stained and mounted for microscopy as reported previously (Halter et al.,1995; Sepp et al.,2000). The rabbit anti-GFP primary was used at 1:200 (AbCAM,Cambridge, UK), mouse anti-β-galactosidase was used at 1:300 (Sigma, St Louis, MO), mouse anti-Neuroglian (mAb 1B7)(Hall and Bieber, 1997) was used at 1:500, rabbit anti-HRP was used at 1:100 (Jackson Immunoresearch, West Grove, PA), mAb 1D4 was used at 1:2, and mAb 22C10 (Developmental Studies Hybridoma Bank, University of Iowa) was used at 1:2. Fluorescent secondaries,goat anti-rabbit Alexa 488 and goat anti-mouse Alexa 568 were used at 1:200(Molecular Probes, Eugene, OR). Images were taken on a BioRad MRC 600 confocal microscope for Fig. 1 andFig. 3B, on a Perkin-Elmer/Yokogawa disk scanning confocal forFig. 3A, and all others were obtained on a BioRad Radiance Plus confocal microscope. Confocal stacks were processed with ImageJ 1.24 and maximum projections were assembled with Adobe Photoshop 5.5.

The in vivo distribution of actin in PNS glial cells over the course of embryonic development has not previously been characterized in any organism. Therefore, before the GTPase mutant analysis was carried out, a confocal analysis of the Actin cytoskeleton in wild-type embryonic glia was performed. For this study, GAL4/UAS system (Brand and Perrimon, 1993) was used to drive the expression of GFP-conjugated actin (UAS-actin-GFP) (Verkhusha et al., 1999) using the repo:GAL4 driver(Sepp et al., 2001), which is expressed in all ectodermally derived embryonic glia. The actin-GFP marker becomes incorporated into filamentous actin within the cells(Verkhusha et al., 1999). Therepo::actin-GFP embryos were visualized using live fluorescence microscopy as well as immunofluorescence microscopy. With these techniques,there were no differences in Actin distribution, and immunofluorescently labeled embryos are presented here.

Peripheral glia arise during early neurodevelopment at the lateral edge of the CNS as a compact cone-shaped mass of cells(Fig. 1A, arrows). Shortly after the pioneer motorneurons from the CNS project axons into the periphery,the peripheral glia, which have proliferated into groups of 6 to 8 cells per hemisegment, begin to extend processes along the newly established axon tracts. The peripheral glia migrate peripherally as a continuous chain of cells (Sepp et al., 2000). With the actin-GFP marker, protrusions were observed which emanated largely from the leading glial cells in each chain(Fig. 1B,C concave arrows). At stage 14, the position of the leading glial cell bodies is variable. In some segments, the cell body, which does not stain darkly with actin-GFP is observed at the leading edge of the glial process(Fig. 1C, solid arrow). In other segments, the cell body is still located among the cone-shaped mass of glia closer to the CNS/PNS transition zone, and in this case, the leading edge of the glial process is often a lamellar-like structure(Fig. 1C, arrowhead). The leading glial cell was identified through comparison of Actin-GFP labeling to previous analyses using glial nuclear and cytoplasmic enhancer traps(Sepp et al., 2000) as well as dye-injection studies (Schmidt et al.,1997). The peripheral glia extend filopodia-like extensions not only along in the correct migratory direction, but also laterally towards the ventral somatic muscle domain (Fig. 1C, upper left concave arrow), suggesting that the glia sample their environment as they travel. The cone-shaped mass of cells begins to resolve such that the ventral peripheral glial cell (vPG;Fig. 1D,E, solid arrows), which will ensheathe the posterior fascicle/segmental nerve (PF/SN), will begin to separate from the rest of the peripheral glia. The majority of peripheral glia ensheathe the anterior fascicle/intersegmental nerve (AF/ISN). As the glial cells advance further along the peripheral nerves, and they expand their cytoplasmic processes to wrap their associated axons, the overall profile of actin becomes smoother (compare Fig. 1D with 1E). Overall, the distribution of actin-GFP is dynamic over the course of glial migration and nerve ensheathement.

The actin-GFP distribution in peripheral glia was compared with the microtubule cytoskeleton in embryos expressing both tau-lacZ(Hidalgo et al., 1995), which binds microtubules, and actin-GFP using the repo:GAL4 driver. The actin-GFP profile of mature embryos appears to have more punctate structures than the long filamentous profile of the microtubule cytoskeleton. The actin-GFP and tau-lacZ patterns both appear fill the same extent of the growing cell processes. For example, the vPG glial processes extend cytoplasmic processes along the SNa motorneuron branch until early larval stages. At the distal tip of this growing sheath, actin-GFP labeling was at the leading edge, as was tau-lacZ staining(Fig. 2A,B, solid arrows). The observation suggests that both the actin and microtubular cytoskeletons are used in peripheral glial process extension. However, from inspection of the overall amount of overlap in staining of actin-GFP and tau-lacZ, the actin and microtubule profiles do not appear to have a large amount of similarity inside the cell processes even though the two patterns are found in all general regions of the cytoplasm (Fig. 2).

To visualize the mature cytoskeletal structure of peripheral nerves better,third instar larvae expressing either actin-GFP or tau-lacZ using therL82#29:GAL4 driver (Sepp and Auld, 1999) were observed. The third instar larva is many times larger than the embryo, thus it is much more amenable to imaging. In live larvae, the actin-GFP profile of the peripheral nerves appeared as a meshwork interspersed with actin-GFP concentrations(Fig. 3A). The punctate concentrations of actin-GFP probably correspond to or are analogous to those seen in late-stage embryos (Fig. 1E). The tau-lacZ profile of stained larval nerves had a long rope-like structure that extended along the length of the glial sheaths towards the first bouton of the axon terminals(Fig. 3B). This pattern is similar to the late embryonic tau-lacZ profile(Fig. 2B). Together, the observations suggest that the actin and microtubule cytoskeletons of the late embryonic glia are resolved into mature networks.

To determine whether the RhoA GTPase is involved in peripheral glial cell migration and axon ensheathement, wild-type and transgenic constructs of Rho were overexpressed in peripheral glia using the repo:GAL4 line. Glial phenotypes were observed using the actin-GFP marker and secondary effects on sensory neurons were observed by double staining embryos with mAb 22C10. Secondary effects on motorneuron development as detected by mAb 1D4 staining were subtle, and are not shown here. It is important to note that expression of the transgenic constructs was generated in a wild-type RhoA background.

Ectopic expression of constitutively active RhoA (RhoAV14) in peripheral glia prevents the cells from migrating peripherally, manifesting as dense clusters of glia arrested at their birthplace at the CNS/PNS transition zone(compare Fig. 1B withFig. 4D). The phenotype of RhoAV14 overexpression is distinctive when compared with all other transgenic phenotypes generated in this study. Typically, long actin-containing fibers extend out of the glial clusters (Fig. 4D, concave arrows), and there are large expanses of PNS tracts with no glial sheaths whatsoever (compareFig. 4D with 4E, solid arrows). The aberrant spike structures of the peripheral glia do not always project along sensory axon pathways (Fig. 4F, concave arrow). The lateral line glia fail to extend processes to interconnect between hemisegments (compareFig. 4A,D, arrowheads) and the lateral chordotonal PNS glial cells appear collapsed and rounded(Fig. 4D, asterisk) although their associated lateral chordotonal neurons appear properly formed(Fig. 4E, asterisk). The sensory axon tracts in these mutants appear defasciculated although their pathfinding to the CNS is generally normal(Fig. 4E, arrow). The glial stalling phenotype is highly penetrant (97%, n=268) compared withrepo::actin-GFP wild types (3%, n=185).

Overexpression of the wild-type form of RhoA caused more subtle glial defects, which included glial cell process stalling(Fig. 4G, solid arrows) as well as incomplete separation of the vPG cell from the other peripheral glia which ensheathe the AF/ISN (Fig. 4G,concave arrow). The overall profile of glial ensheathment of axon tracts does not look entirely normal. Furthermore, defasciculation of the sensory neurons(Fig. 4H, arrow) suggests that the glial sheaths do not correctly wrap the peripheral nerves as in the wild type. The glial stalling phenotype was scored at 23% (n=239).

Similar to overexpression of the wild-type form, the ectopic expression of dominant-negative RhoAN19 yielded subtle glial phenotypes with sensory axon defasciculation as a secondary effect. The vPG cell failed to separate from the main peripheral nerve branch (Fig. 4J, concave arrow) at 26% (n=196) compared with 9%(n=185) in the repo::actin-GFP wild type. All RhoAN19 embryos died before the first larval stage. It has been previously shown that even minor disruption of the peripheral glial-based blood-nerve-barrier can result in embryonic lethality (Auld et al.,1995).

The RhoA embryonic lethal hypomorphic mutants RhoA^E3.10and RhoA^k02107b were analyzed for peripheral glial phenotypes. The mutants were stained with anti-Neuroglian, which recognizes peripheral glial membranes and the epidermis(Fig. 5, green), and the embryos were counterstained with the anti-HRP neuronal marker(Fig. 5, red). The embryonic phenotypes were severe, as RhoA is required for many aspects of embryo development (Magie et al.,1999). In the PNS, HRP-positive sensory neurons were largely absent; however, some motor axon tracts were visible(Fig. 5B). Peripheral glial sheaths were not visible on significant regions of the axon tracts(Fig. 5B, arrows) in all hemisegments analyzed (n=89). As both the RhoA hypomorphic and transgene overexpression phenotypes both showed glial disruption and embryonic lethality, the data suggest that RhoA GTPase is an important mediator of glial migration and axon wrapping.

Activity of the Rac1 GTPase was altered in PNS glia to determine if it also has a role in generation of the nerve sheath. Activity of transgenic Rac1 constructs was tested in a repo::actin-GFP background to observe glial sheaths and embryos were counterstained for sensory neuron phenotypes as in the RhoA experiments above.

Expression of constitutively active Rac1 (Rac1V12) disrupted peripheral glial migration along axon tracts, resulting in unensheathed expanses of neurons (Fig. 6D, solid arrows). In addition, the peripheral glial processes appeared abnormally thin along significant sections of the nerves(Fig. 6D, concave arrows). The peripheral nerves of these embryos were defasciculated(Fig. 6E, arrows). The peripheral glial stall phenotype was scored at 91% (n=257).

When the wild-type form of Rac1 GTPase was overexpressed, the peripheral glia were largely able to migrate into the periphery. However, the overall wrapping profile of peripheral nerves was abnormal, often appearing slightly wider when compared with wild types (Fig. 6A,G, concave arrow). The lateral line glia often did not connect with each other from segment to segment(Fig. 6G, arrowhead). There was a mild degree of sensory axon defasciculation with this mutant(Fig. 6H), where 26%(n=181) of hemisegments showed defasciculation compared to 5% in wild type repo::actin-GFP embryos (n=185).

Similar to the ectopic wild type Rac1 mutant, glial expression of the dominant negative Rac1N17 form generated subtle glial wrapping phenotypes,where the overall profile of the glial sheath was abnormal while the sensory axons were defasciculated (Fig. 6M,N). In some hemisegments (17%, n=163), small gaps in the glial sheath were apparent (Fig. 5M).

Glial expression of the Rac1L89 mutant transgene caused the projection of ectopic lamellar-like structures from the glial sheaths(Fig. 6J, concave arrows). The sensory neuron projections in these mutants were largely normal, but in cases where sensory neurons were misplaced, the peripheral glia extended processes to associate with the aberrant projections(Fig. 6J,K, concave arrows). The sensory axon tracts were defasciculated in many hemisegments (69%,n=194). To examine more closely the ectopic lamellar-like projections of the peripheral glia, embryos with Rac1L89 overexpression were co-labeled for either motor- or sensory neurons. The glial ectopic lamellar-like structures, which varied in size, did not reach out over peripheral motor or sensory branches in many cases (Fig. 7). Using higher magnification, small filopodia-like structures were observed extending from the lamellae(Fig. 7C).

Rac1V12 expression caused the converse effect on glial process extension compared with Rac1L89. The question of whether the absence or collapse of glial sheaths could affect axonal pathfinding was addressed. Surprisingly, the axon pathfinding of both sensory and motoneurons was largely normal(Fig. 8). The most notable observation from the Rac1V12 ectopic expression embryos was that the sensory neurons were more susceptible to defasciculation than the motor tracts.

To explore further whether Rac activity is necessary for peripheral glial development, loss-of-function alleles of Rac1 and the related genesRac2 and Mtl, were analyzed for their PNS phenotypes. Embryos were labeled with anti-Neuroglian and anti-HRP as for the RhoA mutant analysis described above. In hypomorphic Rac1^J10 and nullRac1^J11 homozygous embryos, peripheral glial sheaths showed incomplete coverage of the peripheral axon tracts(Fig. 5C; arrow) in 21%(n=103) and 76% (n=94) of hemisegments, respectively. InRac2^Δ mutants there was incomplete extension of peripheral glial sheaths along the PNS axon tracts(Fig. 5E; arrow) in 18%(n=108) of hemisegments, although the PNS glial sheaths ofMtl^Δ homozygous mutants were largely normal(Fig. 5F). The Rac1and Rac2 mutants also showed neuronal defects as reported previously(Hakeda-Suzuki et al., 2002;Ng et al., 2002). Taken together with the Rac1 transgene overexpression studies, the results suggest that Rac1 is a mediator of glial migration and ensheathment.

To determine whether the GTPase Cdc42 is involved in aspects of peripheral glial development, the constitutively active transgenic constructUAS-Dcdc42V12 and the dominant-negative form UAS-Dcdc42N17were crossed individually to the glial-specific repo:GAL4 driver with the UAS-actin-GFP marker in the background. The glial phenotypes in the resultant progeny were analyzed with live GFP fluorescence and also with immunofluorescence for confocal microscopy. Overexpression of the activated form (Cdc42V12) did not affect peripheral glial migration or nerve wrapping compared with the wild type (data not shown). Furthermore, the embryos were viable and hatched into first instars, suggesting that the seal of the blood/nerve barrier was intact. Overexpression of the dominant-negative form(Cdc42N17) similarly did not lead to any detectable disruptions of the glial sheath or the underlying neuronal patterns (data not shown). Together with previous observations showing that the Cdc42 null mutant has normal sensory neuron patterning (Genova et al.,2000), the data suggest that Cdc42 does not have a major role in peripheral glial development.

Through targeted disruption of GTPase activity, we have shown that Rac1 and RhoA have distinct roles in PNS glial migration and nerve ensheathement,whereas Cdc42 has no apparent function in peripheral glial development. This is the first mutational analysis investigating how peripheral glial cells migrate from the CNS. As well, this study is the first in vivo analysis of peripheral glial actin distribution during migration. We have demonstrated that Drosophila has significant advantages for studying glial cell migration, and this species will most likely be instrumental in the advancing the study of PNS glial development in the future.

For the current study, the actin-GFP was used to image the actin distributions of peripheral glia in the embryonic and larval stages. The transgenic method of labeling actin used in this study most probably represents the endogenous actin distributions in the cell since actin networks are composed of polymerized monomeric actin, and the actin-GFP is incorporated into the intrinsic filamentous actin networks of cells(Verkhusha et al., 1999). This method is currently the most effective means of visualizing peripheral glia in vivo. Standard labeling of embryos with fluorophore-conjugated phalloidin to visualize actin distributions in peripheral glia is not useful, as actin is strongly expressed throughout the embryo, including the somatic musculature. With phalloidin labeling, the actin staining of peripheral glia does not stand out in comparison with the high intensity labeling of the muscles (K.J.S. and V.J.A., unpublished).

Our results showed that the distribution of actin over the course of embryonic development is dynamic, especially during the migratory stages of peripheral glia. As observed previously, peripheral glia migrate into the periphery as a continuous chain of cells(Sepp et al., 2000). With actin-GFP labeling, we observed spike-shaped filopodial protrusions from the leading glia, while the follower glia had a smoother actin distribution. It is possible that the leading glial cell does the majority of pathfinding into the periphery, while the following glia simply adhere to the leading cell during migration. Follower glia must have important rearrangements of their actin, as they all expand their cytoplasmic processes as they migrate peripherally. It is also interesting to note that both actin and microtubules were observed in the region of the leading edge of expanding peripheral glial processes. Thus as for other cells (reviewed by Goode et al., 2000), both microtubules and actin could work together in glia during migration.

The observation that leading glia project spike-shaped protrusions while followers do not is very similar to what has been observed with GFP labeling of lateral line PNS glia in the developing zebrafish(Gilmour et al., 2002), and also for tracheal cells in Drosophila(Ribeiro et al., 2002). It is possible that migrating chains of many different cell types share similar mechanisms for movement. As Drosophila peripheral glial migration is similar to that of zebrafish PNS glia, the data suggest that mechanisms of PNS glial migration could be conserved through evolution. The in vivo distribution of actin in migrating PNS glia in vertebrates is not known. The analysis of actin and microtubules in primary cultures of vertebrate glia is quite different from what we observed. For example, in migrating astrocyte monolayers, actin is observed close to the cell body while microtubules are observed at the extreme leading edge(Etienne-Manneville and Hall,2001). It is likely that the actin distribution of glia varies depending on whether the glial cells contact neurons or not, as nerve ensheathement is affected by actin dynamics, as observed in our experiments. Despite the difficulties in visualizing actin in vivo, it is likely that research in a genetic vertebrate model amenable to imaging such as zebrafish should yield high quality images of glial actin in the future.

The small GTPases Rho, Rac and Cdc42 are expressed in the nervous systems of both Drosophila and vertebrates(Luo et al., 1994;Terashima et al., 2001). The specific functions of the GTPases in axon outgrowth has been studied extensively, although little is known about how they are involved in glial development. In the developing Drosophila embryo, mutations in the small GTPases disrupt the development of many tissues including neurons. This complicates the analysis of GTPase function in glia, as peripheral glia use axonal pathways as a migrational substrate. In Rac GTPase mutations, for example, axon extension is disrupted (Luo et al., 1994; Kaufmann et al.,1998; Hakeda-Suzuki et al.,2002; Ng et al.,2002). Thus, glial stalling phenotypes could not be attributed specifically to loss of GTPase function in the glia. As well, RhoA null mutant embryos lacking maternal contributions of RhoA have severely disrupted overall morphology and segmentation (Magie et al.,1999). Thus, phenotypes in the nervous system would be secondary effects of disrupted embryonic morphogenesis. Complementary analysis of both transgene overexpression and standard loss-of-function mutant analysis is therefore required to understand the specific roles of RhoA, Rac1 and Cdc42 in peripheral glia.

The migration of glia is very different from that of a neuronal growth cone. For glial migration, the cell body containing the nucleus moves and cytoplasmic processes elongate as the glia travel along the axons. When the glia have finished migrating, they must wrap their processes tightly around the axonal bundles to seal the nerves from the surrounding environment. This migration profile is thus unique and is important to understand given that migration and wrapping of glia is implicated in neuropathies and regeneration after nervous system injury.

Our data suggest that RhoA and Rac1 are both involved in peripheral glial cell migration and nerve ensheathement, and have distinct effects on Actin rearrangement. For example, constitutively active Rac1 (V12) and RhoA (V14)expression resulted in halted migration of cell bodies as well as disrupted cytoplasmic process extension. The phenotypes of the two mutants were very different from one another. Rac1 (V12) mutants showed ball-shaped collapsed glia, while RhoA (V14) mutants had very long, spike-shaped actin processes emanating from the cell bodies. The distinct and extreme phenotypes from these mutants suggest that there is a balance of RhoA and Rac1 activity in wild-type peripheral glia to generate normal migration and cytoplasmic process extension. The concept of a balance of GTPase function being necessary for glial cell migration is also supported by our observations that glial cell migration is stalled in both the gain-of-function and loss-of-function mutations. We interpret these observations as suggesting that there is a balance of GTPase activities that is necessary for glial cell migration. In other words, anything that affects this balance either through a loss of function or gain of function, affects the ability of glial cells to migrate.

The well-characterized cultured fibroblast model has shown that Rac is involved in lamellipodia formation, while Rho mediates stress fiber polymerization and Cdc42 is involved in the extension of filopodia(Nobes and Hall, 1995). It is possible that Rac1 and RhoA mediate the assembly of similar structures in peripheral glia. The long, straight actin fibers seen in constitutively active RhoA (V14) mutants could represent overextended stress fibers. Furthermore,the massive glial lamellar-like structures that are stimulated by Rac1L89 expression appear very similar to the lamellipodia of cultured fibroblasts(Nobes and Hall, 1995). The biochemical activity of the Rac1L89 mutation is not known, and can act as either a dominant-negative or constitutively active form, depending on the cell type (Luo et al., 1994;Kaufmann et al., 1998). The Rac1L89 phenotype in peripheral glia was most similar to overexpression of wild-type Rac1 (compare Fig. 5G with 5J), suggesting that the ectopic lamellar structures were a result of moderate increase in Rac1 activity. Thus, it is possible that the Rac1L89 mutation caused Rac1 to be overactive but not as much as in the Rac1V12 mutation.

It was interesting to note that the ectopic actin-containing projections of RhoAV14 and Rac1L89 mutants did not always reach over axon tracts, which are the normal peripheral glial migrational substrates in the wild type. For the steering of a migrating cell, large amounts of actin polymerization occur at the contact between the leading edge of the cell and the attractive migrational substrate (O'Connor and Bentley, 1993; Lin and Forscher, 1993). Perhaps the hyperactivity of the mutant GTPases enabled the peripheral glia to extend processes out on less adhesive substrates compared with axons. It was also interesting to note that ectopic projections of peripheral glia (in the RhoAV14 and Rac1L89 mutants) did not interfere with axon pathfinding in the periphery. The ectopic glial projections could be a result of failed glial pathfinding instead. Interestingly, axons were capable of correctly migrating in the absence of glial sheaths (in the RhoAV14 and Rac1V12 mutants). It has been noted previously that peripheral glia mediate sensory axon guidance to the CNS(Sepp et al., 2001). Thus,peripheral glia most probably mediate sensory axon migration to the CNS using secreted cues.

In conclusion, we have shown that the rearrangement of actin by the small GTPases RhoA and Rac1 is an integral part of peripheral glial migration and nerve ensheathement. This model can be used in the future to dissect GTPase signaling pathways. For example, it could be possible to determine whether the GTPases share common activators or effectors given that the cellular phenotype for mutants of each GTPase is so distinctive. In addition, we have found further characteristics of Drosophila peripheral glial development that are similar to the vertebrate zebrafish model. Thus, future genetic research in Drosophila to understand PNS glial migration will probably help advance the knowledge of the vertebrate field as well.

We thank Andrea Brand, Nick Harden, Michael Hortsch, Bill Leiserson, Liqun Luo, Marek Mlodzik, Hiroki Oda, the Bloomington Stock Center and the Developmental Studies Hybridoma Bank for fly stocks and antibodies. We also sincerely thank Nick Harden and Tim O'Connor for helpful discussions and comments on the manuscript. K.J.S. is supported by a Rick Hansen Neurotrauma Initiative fellowship. The work was supported by grants from the Natural Science and Engineering Research Council of Canada and the Canadian Institutes of Health Research.