The actin-binding protein UNC-115 is an effector of Rac signaling during axon pathfinding in C. elegans

Rac GTPases, members of the Rho-family of small GTPases that also includes Rho and Cdc42, are central regulators of cellular morphogenesis, including axon pathfinding (reviewed by Dickson,2001; Hall, 1998;Luo, 2000;Zigmond, 1996). Three C. elegans Racs, CED-10, MIG-2 and RAC-2/3, have overlapping roles in axon pathfinding and cell migration (Lundquist et al., 2001; Wu et al.,2002) and ced-10 and mig-2 are required redundantly for vulval epithelial morphogenesis(Kishore and Sundaram, 2002)and P-cell migration (Wu et al.,2002). Rac redundancy is conserved in Drosophila, where three Racs (Rac1, Rac2 and Mtl) have overlapping functions in axon pathfinding, myoblast fusion and epithelial morphogenesis(Hakeda-Suzuki et al., 2002;Ng et al., 2002). Although Rac function is redundant in some processes, C. elegans Racs have individual requirements in other processes. CED-10 is required for phagocytosis of cells undergoing programmed cell death and for the migration of the distal tip cells of the hermaphrodite gonad(Reddien and Horvitz, 2000),and MIG-2 is required for the migration of the Q neuroblast descendants and the distal tip cells (Zipkin et al.,1997). Redundancy of Rac function is likely to be conserved in vertebrates, which also have multiple Rac genes(Didsbury et al., 1989;Haataja et al., 1997). However, mouse Rac1 is required individually for gastrulation and cell migration during development(Sugihara et al., 1998) and mouse Rac2 is required for neutrophil migration(Roberts et al., 1999). Thus,metazoan genomes contain multiple Rac genes that have distinct as well as overlapping roles in a variety of morphogenetic events in vivo. Despite redundancy of Rac function in axon development, gain-of-function constitutively active and dominant-negative Racs perturb axon pathfinding and cell migration in C. elegans, Drosophila and mouse(Lundquist et al., 2001;Luo et al., 1996;Luo et al., 1994;Zipkin et al., 1997).

Like other Ras superfamily GTPases, Racs cycle between a GTP-bound active state and a GDP-bound inactive state(Bourne et al., 1991). Molecules that regulate Rho GTPase activity include GTP exchange factors(GEFs), which stimulate exchange of GDP for GTP, promoting the active state(Kaibuchi et al., 1999). A number of dbl-homology (DH) GEFs have been implicated in controlling Rac activity, including UNC-73 Trio (Bellanger et al., 1998; Steven et al.,1998), which acts with each of the three Racs in C. elegans and Drosophila axon pathfinding in vivo(Awasaki et al., 2000;Bateman et al., 2000;Liebl et al., 2000;Lundquist et al., 2001;Newsome et al., 2000).

In the developing nervous system, growth cones, which are dynamic sensorimotor structures at the distal tip of extending axons, sense and respond to extracellular guidance cues by modulating their actin cytoskeletons, resulting in changes in direction of outgrowth(Tessier-Lavigne and Goodman,1996). Growth cones are composed of lamellipodia and filopodia,which are dynamic, actin-based plasma membrane extensions that underlie outgrowth and guidance (Letourneau,1996). Rac GTPases control cell shape at least in part by regulating the structure and dynamics of the actin cytoskeleton(Hall, 1998) and Rac activity can induce lamellipodia in cultured cells(Ridley et al., 1992). Recent studies have identified a number of downstream Rac effector molecules that adapt Rac activity to the actin cytoskeleton(Bishop and Hall, 2000), some of which involve activation of the actin-nucleating Arp2/3 complex. For example, Rac interacts with the adapter protein IRSp53 which in turn links the Arp2/3 potentiator WAVE2 (Takenawa and Miki, 2001), and Rac1 along with the adapter protein Nck acitvate the Arp2/3 potentiator Scar/WAVE (Eden et al., 2002). Furthermore, the actin-interacting protein cortactin,also an Arp2/3 activator (Weaver et al.,2001), has been implicated as downstream Rac effector(Weed et al., 1998). Distinct mechanisms involve Rac interaction with p21-activated kinase, which in turn can influence the cytoskeletal effector merlin(Kissil et al., 2002;Xiao et al., 2002), and potential Rac regulation of the actin-binding protein gelsolin(Azuma et al., 1998). Thus, Rac activity can influence the actin cytoskeleton by a variety of pathways,possibly reflecting the multiple and overlapping functions of Racs in development. Undoubtedly, many downstream cytoskeletal effectors of Racs remain to be identified. Rac activity in different morphogenetic events might be controlled by different combinations of upstream Rac regulators (GEFs) and downstream cytoskeletal effectors. Mechanisms by which multiple Rac regulators, Racs and downstream effectors control different morphogenetic events in vivo remain unclear.

In this work we provide evidence that UNC-115 is a downstream Rac effector in C. elegans axon pathfinding. UNC-115 is similar to the human actin-binding protein abLIM/limatin (Roof et al., 1997) and is involved in axon pathfinding(Lundquist et al., 1998). We show that UNC-115 acts with the Racs and UNC-73/Trio GEF in axon pathfinding:UNC-115 acts in parallel to CED-10 and MIG-2, possibly in the RAC-2/3 pathway,and UNC-115 is required for the effects of constitutively active RAC-2 on axon morphogenesis. Furthermore, we show that UNC-115 is an actin-binding protein,suggesting that UNC-115 is a new downstream cytoskeletal effector of Rac signaling that acts specifically in the RAC-2 branch of the triply-redundant Rac pathway in axon pathfinding.

Nematodes were cultured by standard techniques(Brenner, 1974;Sulston and Hodgkin, 1988). All experiments were performed at 20°C. The following mutations were used.

LGI: unc-73(e936, rh40), kyIs5[ceh-23::gfp,lin-15(+)]

LGIV: ced-10(n1993), dpy-13(e184), lqIs3[osm-6::gfp,lin-15(+)]

LGX: unc-115(mn481, ky275), mig-2(mu28, rh17, gm38 mu133),lin-15(n765), lqIs2[osm-6::gfp, lin-15(+)],lqIs10[ceh-10::gfp, lin-15(+)],kyIs4[ceh-23::gfp, lin-15(+)]

Germline transformation of nematodes was performed by standard techniques(Mello and Fire, 1995) usinglin-15(n765) mutants and lin-15(+) DNA(Clark et al., 1994) as a marker for co-transformation. Extrachromosomal arrays were integrated into the genome using trimethylpsoralen/UV mutagenesis(Anderson, 1995) and standard techniques (Mello and Fire,1995). RNAi was performed by dsRNA injection (1 μg/μl) into the body cavity or gonad as described(Fire et al., 1998).

CAN axon morphology and cell position was scored using the integratedceh-23::gfp transgenes kyIs4 X and kyIs5 I, or the integrated ceh-10::gfp transgene lqIs10 X, as described(Lundquist et al., 2001). A posterior CAN axon was scored as mutant if, because of premature termination or misguidance, the axon failed to extend further than half the distance between the vulva and the phasmid neurons in the tail (approximately to the region of the postdeirid ganglion). CAN cell migration was scored as mutant if the CAN cell body was greater than two CAN cell body-widths from the vulva.

PDE morphology was examined in animals harboring the integratedosm-6::gfp transgenes lqIs3 IV and lqIs2 X that were expressed in all ciliated sensory neurons, including PDE. A PDE axon was scored as having a pathfinding defect if, because of premature termination or axon wandering, the axon failed to reach the VNC. A PDE axon was scored as exhibiting ectopic axons if one or more ectopic axons were seen emanating from the normal axon or from the PDE cell body. A PDE cell was scored as having abnormal cellular morphology if sheet-like membrane extensions and/or finger-like membrane extensions were observed anywhere on the cell, including on the axons, dendrites and cell bodies. Each genotype was scored on at least three separate occasions, and >100 animals were scored. Twice the standard error of the proportion (percentage) is displayed in the tables. Genetic interactions were considered synthetic (the genes have redundant, overlapping functions) when the proportion of defects in the double mutant was greater than the additive effects of each mutant alone.

Standard molecular biological techniques were used(Ausubel et al., 1987;Sambrook, 1989). Oligonucleotide primer sequences used for PCR are available upon request. The sequences of all coding regions of clones derived using PCR in this work were determined to ensure that no errors were introduced. Theosm-6::gfp transgene was produced by fusing the osm-6promoter (bases 2680-2373 of cosmid F58H1) produced by PCR from C. elegans genomic DNA upstream of gfp. osm-6::gfp transgenic lines were constructed by microinjection at a concentration of 30 ng/μl.

Each rac-coding region, including initiator ATG, introns and stop codon, was amplified by PCR from genomic DNA: ced-10 (bases 36,193-34,079 of cosmid C09G12); mig-2 (bases 20,146-21,722 of cosmid C35C5); and rac-2 (bases 105-1957 of cosmid K03D3). Each coding region was placed downstream of the osm-6 promoter and theunc-115 promoter. The G12V mutation (G16V in mig-2) was introduced into the ced-10-, mig-2- and rac-2-coding regions using the Quikchange Site-Directed Mutagenesis System (Stratagene, La Jolla,CA). Rac transgenes were injected in germline transformation experiments at 5 ng/μl, and the osm-6::gfp plasmid was co-injected at 30 ng/μl. Multiple transformed lines were obtained for each construct(>3). Each line displayed similar behavior, and representative lines from each transgene are shown in Table 3.

To score the effects of unc-115 mutation on PDE axons harboringrac transgenes, arrays were crossed into the unc-115(ky275)mutant background. At least two independent arrays consisting of each transgene were scored in the unc-115 background with similar results. One representative line from each is shown inTable 3. Furthermore, the transgenes were re-isolated by outcrossing from unc-115 and rescored. PDE defects were restored to a degree similar to what was observed before building the unc-115; rac-2(G12V) double (data not shown).

The gfp-coding region was placed upstream of and in frame with therac-2-coding region in the unc-115 promoter::rac-2(+) andunc-115 promoter::rac-2(G12V) transgenes. Frameshifts betweengfp- and rac-2-coding regions were produced by fusinggfp upstream of and out of frame with the rac-2-coding regions. unc-115::gfp::rac-2 transgenes were microinjected at concentrations of 5 ng/μl to generate extrachromosomal arrays.

The program MODELLER was used to model the structure of the UNC-115 VHD from the chicken villin VHD (HP67; 1qqv.pdb). Charge distribution was determined using SYBYL program.

For actin sedimentation assays, a fragment of the C-terminal region of theunc-115 cDNA, including the VHD-coding region (from position 1350 in F09B9.2b open reading frame sequence to the stop codon), was fused in frame to the 6-histidine moeity and DHFR in the pQE-41 vector (Qiagen, Valencia, CA). Point mutations in the VHD were constructed using the Quikchange Site-Directed Mutagenesis System (Stratagene, La Jolla, CA). The resulting 6HIS::DHFR::UNC-115 proteins were purified by standard Ni⁺⁺chelation chromatography (Qiagen, Valencia, CA). Actin co-sedimentation assays were based on standard procedures (Miller et al., 1991): 5 μM 6HIS::DHFR::UNC-115 protein with and without 20 μM rabbit skeletal muscle G-actin was mixed in 5 mM Tris pH 7.5,0.2 mM ATP, 0.2 mM DTT, 0.2 mM CaCl₂ and actin was polymerized by adding 20 mM MgCl₂, 5 mM ATP, 100 mM KCl and incubating at room temperature for 1 hour. Actin filaments were sedimented by centrifugation(130,000 g for 40 minutes). The pellet was analyzed by SDS-PAGE and western blotting with anti-RGS-6HIS antibody (Qiagen, Valencia,CA).

We have shown previously that the three C. elegans rac genesced-10, mig-2 and rac-2/3 have overlapping functions in CAN axon and D-class motor axon development(Lundquist et al., 2001). To determine if rac redundancy is conserved in other neurons, we analyzed the effects of racs on the development of the axons of the PDE neurons. The PDE neuron is also a useful model for studying axon pathfinding: PDE axon morphology is relatively simple; the PDE axons can be unambiguously distinguished from the axons of other neurons; and PDE function is not required for viability. The two PDEs are a bilaterally-symmetric pair of ciliated neurons that reside at the lateral midline in the post-deirid ganglion in the posterior of the animal(Fig. 1A)(White et al., 1986). A single, unbranched axon from the cell body extends straight ventrally to the ventral nerve cord (VNC) (Fig. 1B), where the axon bifurcates and extends anteriorly and posteriorly in a fascicle with the other PDE axon in the VNC(Fig. 1F). A ciliated dendrite that is exposed to the external environment via a pore in the hypodermis and cuticle extends dorsally from the PDE cell body(Fig. 1B).

We analyzed PDE axon morphology in rac single and double loss-of-function mutants. rac-2/3 activity was perturbed using RNAi(Fire et al., 1998). Therac-2/3 locus is composed of two nearly identical rac genes,and rac-2/3(RNAi) is predicted to reduce the function of both genes(see Lundquist et al., 2001).ced-10, mig-2 and rac-2/3(RNAi) single mutants displayed few defects in PDE development (Table 1). Each pairwise double mutant combination of ced-10(n1993),mig-2(mu28) and rac-2/3(RNAi) displayed synthetic PDE axon defects (Table 1), including axon guidance defects (the axons failed to reach the VNC and often wandered along the lateral body wall) (Fig. 1C,D) and premature axon termination and defasciculation in the ventral nerve cord (Fig. 1E,F). In addition to these defects in axon pathfinding, rac double mutants displayed ectopic axon formation (ectopic axons formed as branches from the main axon or emanated from the cell body)(Fig. 1G). Thus, racgenes control multiple aspects of PDE axon development, including axon pathfinding (guidance, outgrowth and fasciculation), as well as suppression of ectopic axon formation. Similar results were obtained using themig-2(gm38mu133) allele (Table 1). Each rac double mutant displayed the entire spectrum of defects, suggesting that all three rac genes act in each process as opposed to individual rac gene involvement in a single aspect of PDE axon development.

ced-10(M+); mig-2 double mutants also displayed defects in PDE dendrite development (data not shown). The PDE dendrite was missing or misshapen (misguided, lacking a discernible cilium or exhibited ectopic branches) in 17% of ced-10(M+); mig-2 PDE neurons.rac-2/3(RNAi) double mutants did not display dendrite defects.

UNC-73 is a Trio-like molecule with two DH-GEF domains, one of which, GEF1,is a Rac-GEF (Steven et al.,1998). unc-73 acts with the three Racs in CAN and D-class axon development (Lundquist et al.,2001), and here we find that unc-73 also affects PDE axon development. The incomplete loss-of-function mutants unc-73(rh40) andunc-73(e936) displayed PDE axon defects similar to racdouble mutants, including axon guidance defects, axon defasciculation and premature termination in the VNC, and ectopic axon formation(Table 1).unc-73(rh40) and unc-73(e936) PDE axon pathfinding defects and ectopic axon formation were enhanced significantly by ced-10,mig-2 and rac-2/3(RNAi)(Table 1), suggesting thatunc-73 and the three rac genes act together in PDE axon development. However, the fact that rac mutations enhanceunc-73(rh40), a mutation that specifically attenuates the GEF1 Rac-GEF activity of UNC-73, raises the possibility that the Racs might act in a pathway parallel to UNC-73. To determine if RAC-2 acts in the UNC-73 pathway, we tested the ability of constitutively-active rac-2(rac-2 harboring the G12V mutation; see below) to suppress PDE axon defects caused by unc-73(rh40). Indeed, transgenic expression of constitutively-active rac-2 in the PDE neuron partially suppressed the PDE ventral axon guidance defects caused by unc-73(rh40)(Table 1). Furthermore,unc-73(rh40) PDE axons often wandered laterally before reaching the VNC, and this wandering was also partially suppressed by activatedrac-2 (data not shown). These data suggest that rac-2 acts downstream of unc-73 in the same pathway. Similar suppression ofunc-73 defects in the D-class motor axons has been observed with constitutively active ced-10 and mig-2(Wu et al., 2002). These results combined with the biochemical data that UNC-73 acts as a GEF on CED-10 and MIG-2 (Wu et al., 2002)strongly suggest that ced-10, mig-2, rac-2/3 and unc-73 act in the same pathway in axon pathfinding. As described by Lundquist et al.(Lundquist et al., 2001),enhancement of unc-73(rh40) axon pathfinding defects by racmutation could be due to Rac regulation by other molecules in addition to UNC-73 or Rac regulation by a domain of UNC-73 apart from the GEF1 Rac-GEF domain.

Although these data indicate that rac-2 acts in theunc-73 pathway, they do not exclude the possibility that UNC-73 has additional, Rac-independent roles in PDE axon development.unc-73(e936) is predicted to affect multiple UNC-73 activities,including that of GEF2 Rho-GEF (Steven et al., 1998). Disruption of a rac-independent activity of UNC-73 by unc-73(e936) might contribute to rac enhancement of unc-73(e936).

Dendrite development was also perturbed by unc-73(rh40 ande936) and these defects were enhanced by ced-10 andmig-2 (data not shown).

unc-115 mutations cause defects in axon pathfinding andunc-115 encodes a putative actin-binding protein(Lundquist et al., 1998). Racs are thought to mediate cellular morphogenesis in part by regulating the structure and dynamics of the actin cytoskeleton(Hall, 1998). To determine ifunc-115 acts with rac genes in axon pathfinding, we analyzed the CAN and PDE axons of double loss-of-function mutants of unc-115with ced-10, mig-2 and rac-2/3(RNAi). Neitherunc-115(ky275), a putative unc-115 null allele, norunc-115(mn481), an incomplete loss-of-function allele, caused defects in CAN axon pathfinding (Table 2) [for a description of CAN axon pathfinding and defects see Lundquist et al. (Lundquist et al.,2001)]. However, unc-115 animals displayed weak PDE axon guidance defects: 12% of unc-115(ky275) PDE axons (n=174)and 6% of unc-115(mn481) PDE axons (n=189) wandered laterally (>45°C from straight ventrally) on their trajectory to the VNC. All unc-115 PDE axons eventually reached the VNC. Furthermore,unc-115 mutants alone displayed ectopic PDE axons (20% of PDEs inunc-115(ky275)) (Table 2). Thus, unc-115 mutations alone had weak effects on PDE axon development, most notably ectopic axon formation.

We found that unc-115 synergized with ced-10 andmig-2 but not rac-2/3 in CAN and PDE axon pathfinding.unc-115; ced-10 and unc-115 mig-2 double mutants were viable and fertile, but displayed synthetic defects, including a withered tail (Wit),similar to but weaker than rac double mutants and unc-73animals. The Wit phenotype is associated with perturbed CAN neuron development(Forrester et al., 1998) and indeed, ced-10; unc-115 and mig-2 unc-115 animals displayed failure of the posterior CAN axon to extend into the tail due to axon wandering or premature termination (Table 2). Furthermore, unc-115; ced-10 and unc-115 mig-2 animals displayed synthetic PDE defects, including ventral axon guidance errors (Table 2) and axon defasciculation in the VNC (data not shown). These data indicate thatunc-115 has overlapping function with ced-10 andmig-2 in CAN and PDE axon pathfinding.

By contrast, the CAN and PDE axons of unc-115; rac-2/3(RNAi)animals resembled those unc-115 alone(Table 2), suggesting thatunc-115 and rac-2/3 affect the same pathway in CAN and PDE axon pathfinding, possibly a pathway in parallel to mig-2 andced-10. However, we cannot rule out the possibility thatunc-115 also acts in the ced-10 pathway, becauseced-10(n1993) is not a null allele(Soto et al., 2002) (E. A. L.,unpublished results).

In addition, we found that unc-115 acts with unc-73 in axon pathfinding. unc-115(ky275) and unc-115(mn481)mutations significantly enhanced the CAN and PDE axon pathfinding defects caused by unc-73(e936) and unc-73(rh40) mutations(Table 2), including CAN posterior guidance errors and PDE ventral guidance errors. PDE defects ofunc-115; unc-73 double mutants were not further enhanced byrac-2/3(RNAi) (Table 2), confirming that unc-115 and rac-2/3 act in the same pathway. Taken together, these results indicate that unc-115acts with the rac genes and unc-73 to mediate CAN and PDE axon pathfinding.

While unc-115 and unc-73 mutants and rac double mutants displayed ectopic axon formation, double mutants of unc-115with ced-10, mig-2 and unc-73 did not display enhanced ectopic axon formation (Table 2). Possibly, unc-115 acts specifically withrac-2/3 in axon pathfinding and with all three rac pathways in the suppression of ectopic axon formation.

The three racs and unc-73 affect other actin-based morphogenetic processes, including migration of the CAN cell body(Lundquist et al., 2001).unc-115 alone or in double mutant combinations with the racsand unc-73 had no effect on CAN cell migration(Table 2). Indeed,unc-115 doubles with ced-10, mig-2 and unc-73showed no enhanced CAN cell migration defect despite enhanced CAN axon pathfinding defects. ced-10, mig-2 and rac-2/3 also control the migrations of the distal tip cells of the hermaphrodite gonad, andced-10 and rac-2/3 are involved in phagocytosis of cells undergoing programmed cell death(Lundquist et al., 2001;Reddien and Horvitz, 2000).unc-115 had no effect on distal tip cell migration or phagocytosis alone or in any double mutant combination with the racs andunc-73 (P. Reddien, personal communication). Thus, unc-115is required for CAN and PDE axon pathfinding but is apparently not involved in CAN and distal tip cell migration and cell corpse phagocytosis.

In order to understand the molecular relationships UNC-115 and the three Racs, we studied the effects of constitutively active Racs on axon pathfinding. The glycine to valine mutation at position 12, which is canonical for constitutive activation of Ras superfamily GTPases, attenuates GTPase activity, thus favoring the GTP-bound, active state(Bourne et al., 1991).mig-2(rh17), an allele with the equivalent G16V activating mutation, disrupts HSN and CAN axon pathfinding, and ced-10(G12V)dominantly perturbs CAN axon pathfinding and cell migration(Lundquist et al., 2001;Zipkin et al., 1997). To test if constitutively active rac-2 can also dominantly perturb axon pathfinding, we constructed a transgene containing the rac-2-coding region harboring the G12V mutation under the control of theosm-6 promoter. The osm-6 promoter is expressed exclusively in ciliated neurons, including PDE, and shows no other discernible expression(Collet et al., 1998). Similar constructs were made using the ced-10(G12V)- andmig-2(G16V)-coding regions [collectively referred to as therac(G12V) transgenes], as well as the wild-type coding region of eachrac. Animals harboring the rac-2(G12V) transgene displayed dominant defects in PDE axon development that resembled the defects ofrac double loss-of-function mutants, including axon guidance errors(Table 3), axon defasciculation and ectopic axon formation (Fig. 2A,B; Table 3).ced-10(G12V), mig-2(G12V) and mig-2(rh17) showed similar defects (Table 3). The pathfinding and fasciculation errors caused by the rac(G12V)transgenes were generally weaker than the equivalent defects caused byrac loss-of-function, whereas ectopic axon formation was generally more severe in rac(G12V) animals than in the racloss-of-function mutants (compare Table 2 with Table 3, andFig. 1E withFig. 2A,B). The PDE dendrite also displayed ectopic branching in each of the three rac(G12V)animals (Fig. 2B). The wild-type versions of each rac transgene caused PDE axon defects similar to but weaker than the rac(G12V) transgenes, most often weak ectopic axon formation (Table 3). rac(G12V) expression driven by a different promoter,the neuron-specific unc-115 promoter, caused similar defects (data not shown).

These data indicate that Rac(G12V) molecules dominantly perturb PDE axon development. Several lines of evidence suggest that rac(G12V)transgenes produced gain-of-function, constitutively active Rac molecules:similar mutations in other Ras superfamily GTPases produce constitutive activation; the effects of rac(G12V) were dominant to wild type; the effects of ced-10(G12V) did not require wild-type ced-10(data not shown); and the effects of rac(G12V) transgenic expression were more severe than rac(+) transgenic expression.

In addition to pathfinding defects and ectopic axon formation,constitutively active racs caused PDE cell shape alterations not seen in rac loss-of-function double mutants. Animals harboring any of the three rac(G12V) transgenes displayed sheet-like extensions of plasma membrane and multiple, thin processes emanating from the cell bodies, axons and dendrites (Fig. 2C,D). Ectopic axons often emanated from the sheet-like extensions. The thin processes were generally thinner and shorter than the normal or ectopic axons and were often observed at the edges of the sheet-like membrane extensions. Despite abnormal cellular morphology, the PDE axons usually completed their extensions to the VNC.

rac(G12V)-induced ectopic extensions displayed dynamic morphology even in adult animals (Fig. 2E,F). For example, the sheet-like structures were observed to form single or multiple neurite-like extensions, and the thin processes were apparently extended and retracted over time. These results demonstrate that constitutively active racs induced dynamic cellular structures,plasma membrane extensions and thin processes, not seen in rac,unc-73 or unc-115 loss-of-function mutants. However, these structures resemble lamellipodia and filopodia normally found on growth cones during axon outgrowth (Knobel et al., 2001). Possibly, Rac(G12V) molecules ectopically induce these morphogenetic structures.

The loss-of-function studies described above place unc-115 in therac-2/3 pathway. To test if UNC-115 acts downstream of RAC-2, we determined if loss of unc-115 function could suppress the effects of constitutively active rac-2(G12V). We found that the null mutationunc-115(ky275) suppressed the dominant effects ofrac-2(G12V) (Table 3). The strong ectopic axon formation induced by rac-2(G12V) was suppressed by unc-115(ky275) to a level similar tounc-115(ky275) alone, and unc-115(ky275) suppressed the ectopic plasma membrane extensions induced by rac-2(G12V). Whereas the axon pathfinding defects caused by rac-2(G12V) were weak (1%), we saw no axon pathfinding defects in unc-115(ky275); rac-2(G12V) double mutants. By contrast, unc-115(ky275) did not suppress the effects of constitutively-active ced-10(G12V), mig-2(G16V) ormig-2(rh17) (Table 3). Instead, unc-115(ky275) slightly enhanced the effects ofced-10(G12V) and mig-2(G16V). These results indicate that UNC-115 activity mediates the effects of constitutively active RAC-2(G12V). Although it is possible that RAC-2(G12V) perturbs a process in which Racs are not normally involved, the loss-of-function data that place rac-2/3and unc-115 in the same pathway combined with unc-115suppression of rac-2(G12V) strongly suggest that UNC-115 acts downstream of RAC-2 in PDE axon development.

Rho GTPases, including Racs, are anchored to the plasma membrane by covalent attachment of prenyl group mediated by a C-terminal CAAX box(Zhang and Casey, 1996), and GFP::MIG-2 and GFP::CED-10 accumulate at the plasma membrane(Lundquist et al., 2001;Zipkin et al., 1997). The potential RAC-2 polypeptide contains a consensus C-terminal CAAX box (CTVL). To determine if RAC-2 accumulates at the plasma membrane, a rac-2transgene was generated that consisted of the wild-type rac-2-coding region fused in-frame downstream of the green fluorescent protein(gfp) (Chalfie et al.,1994) coding region (Fig. 3A). This transgene is predicted to encode a full-length RAC-2 molecule tagged with GFP at the N terminus. The expression of thegfp::rac-2 transgene was placed under the control of theunc-115 promoter, which is expressed in most if not all neurons and neuroblasts (Lundquist et al.,1998). Animals harboring the unc-115::gfp::rac-2transgene showed expression of GFP::RAC-2 that accumulated at the cell margins of neuroblasts and neurons as well as in the nerve ring(Fig. 3B-D). The nerve ring is composed of the axons of many neurons(White et al., 1986), and nerve ring accumulation suggests that GFP::RAC-2 localized to the plasma membranes of axons. GFP::CED-10 and GFP::MIG-2 show similar plasma membrane and nerve ring accumulation (Lundquist et al., 2001; Zipkin et al.,1997). Wild-type rac-2(Fig. 3) andrac-2(G12V)-coding regions displayed indistinguishable localization,and gfp::rac-2(G12V) caused dominant effects on PDE development as described above (data not shown). Expression of GFP from a transgene that contained a frameshift between the gfp- and rac-2-coding regions failed to accumulate at the cell margins and in the nerve ring(Fig. 3E-G) and instead was uniformly distributed in the cytoplasm, indicating that RAC-2 sequences mediate accumulation of GFP::RAC-2 at the plasma membrane, presumably via C-terminal CAAX motif.

The predicted UNC-115 polypeptide consists of three N-terminal LIM domains and a C-terminal region similar to the actin-binding headpiece domain of villin (VHD) (Fig. 4A,B)(Lundquist et al., 1998). Not all VHDs can bind to actin, and those VHDs with demonstrated actin-binding ability have conserved basic residues that form a `positive patch' that might mediate molecular contacts with actin(Vardar et al., 2002). We modeled the structure of the UNC-115 VHD based upon the NMR structure of the chicken villin VHD (called HP67) (Fig. 4B,C). The UNC-115 VHD had hallmark features of all VHDs,including a hydrophobic `cap' and a charged `crown'(Fig. 4C) (seeVardar et al., 2002). Furthermore, the UNC-115 VHD had a prominent `positive patch' similar to other actin-binding VHDs. We next tested the ability of the UNC-115 VHD to co-sediment with F-actin in vitro. A bacterially expressed fragment of UNC-115 containing the VHD sedimented in the presence of but not in the absence of actin filaments (Fig. 4D). We generated a mutant form of the UNC-115 VHD in which basic residues that contribute to the `positive patch' were changed to acidic residues(Fig. 4B,C). This mutant VHD failed to co-sediment with actin filaments(Fig. 4D). These results demonstrate that the UNC-115 VHD is an actin filament binding domain.

While some downstream effectors of Rac signaling in morphogenesis are known, many remain to be identified. Furthermore, questions remain about how multiple Racs and effector molecules are deployed during development to mediate specific morphogenetic events such as axon pathfinding. We have identified and characterized UNC-115 as a new downstream cytoskeletal effector of Rac signaling in axon pathfinding.

Three C. elegans rac genes define three overlapping pathways that regulate the development of CAN and D-class axons, and unc-73 Trio GEF is likely to control all three rac pathways in this process(Wu et al., 2002;Lundquist et al., 2001). Furthermore, UNC-73 Trio is known to act as a GEF for multiple Racs, including the canonical Racs and the MIG-2-like Racs(Wu et al., 2002;Newsome et al., 2000;Steven et al., 1998). We have shown that the three racs act redundantly in PDE axon development. Multiple aspects of PDE development are affected by racloss-of-function, including axon guidance, axon fasciculation and axon outgrowth as well as suppression of ectopic axon formation, indicating thatrac genes control multiple aspects of axon development. We also provide evidence that the three racs act with unc-73 in axon pathfinding: each rac loss-of-function mutation enhanced weakunc-73 mutations and rac-2(G12V) suppressedunc-73(rh40). Together with results showing that UNC-73 GEF1 acts as a GEF on CED-10 and MIG-2 (Wu et al.,2002), these data indicate that the racs andunc-73 act in the same pathway in PDE axon pathfinding.

UNC-115 is a molecule similar to the human actin-binding protein abLIM/limatin (Lundquist et al.,1998; Roof et al.,1997) and consists of three N-terminal LIM domains, which are thought to mediate protein-protein interactions(Dawid et al., 1998), and a C-terminal villin headpiece domain, an actin-binding domain found in a variety of proteins (Vardar et al.,2002). UNC-115 is required for pathfinding of many but not all axons in C. elegans (Lundquist et al., 1998), and dominant-negative abLIM/limatin can perturb RGC axon pathfinding in the developing mouse visual system(Erkman et al., 2000). UNC-115 might act as a cytoskeletal adapter protein that interacts with actin via the VHD and with other molecules via the LIM domains. We show that UNC-115 acts downstream of Rac signaling during axon pathfinding, suggesting that UNC-115 might adapt Rac activity to the growth cone actin cytoskeleton.

In C. elegans, unc-115 is expressed in most if not all neurons throughout development, yet many neurons display normal or near-normal axon pathfinding and development in unc-115 null mutants(Lundquist et al., 1998). Our results explain this paradoxical observation by demonstrating thatunc-115 acts in the rac-2/3 branch of the tripartiterac cascade and has overlapping function with the ced-10 andmig-2 pathways in axon pathfinding. However, ced-10(n1993)is not a null allele and unc-115 might enhanceced-10(n1993). Therefore, unc-115 might act in both therac-2/3 and ced-10 pathways. If this is the case, then there must be other genes that act in parallel to unc-115 in theced-10 pathway, as the unc-115 null phenotype is viable and fertile and does not resemble the ced-10 null. Althoughunc-115 might act in the rac-2/3 pathway in axon pathfinding, unc-115 might act with all three racs in the suppression of ectopic axons, as unc-115 mutants alone display ectopic axon formation that is not enhanced by mutations in the threerac genes.

The genetic relationships between the rac genes, unc-73and unc-115 in axon pathfinding are shown inFig. 5. Although our results indicate that UNC-73 controls the three Racs in axon development, they do not exclude the possibility that UNC-73 has additional, Rac-independent effects on axon pathfinding. In addition to the GEF1 Rac GEF domain, UNC-73 has a second GEF domain (GEF2) that acts on the Rac-related small GTPase Rho(Spencer et al., 2001;Steven et al., 1998). Possibly, rho-1, the single C. elegans gene that encodes Rho, acts downstream of unc-73 in parallel to the racs to control PDE axon development.

We found that unc-115 was not required for other morphogenetic events involving the racs and unc-73, including migration of the CAN cell bodies, migration of the distal tip cell of the hermaphrodite gonad and phagocytosis of cell corpses undergoing programmed cell death.unc-115 might be a rac effector that acts specifically in axon pathfinding. Possibly, other genes with roles similar to unc-115act with the racs and unc-73 in other morphogenetic events.

PDE neurons of animals harboring ced-10(G12V), mig-2(G16V) andrac-2(G12V) constitutively-active transgenes displayed ectopic structures not seen in rac loss-of-function mutants, including extensive networks of ectopic axons. rac(G12V) animals also displayed ectopic plasma membrane extensions consisting of large, sheet-like structures,as well as thin, finger-like projections emanating from cell bodies, axons and dendrites. These structures were dynamic over time, even in adult animals. Although we have no direct evidence of the nature of these structures, their form and dynamics resemble actin cytoskeleton-based lamellipodia and filopodia found in growth cones of developing animals(Knobel et al., 1999). Rac activity is known to induce ectopic lamellipodia in cultured cells(Ridley et al., 1992),consistent with our observation of lamellipodia-like structures induced by Rac activity in vivo. However, we also observed filopodia-like extensions induced by Rac activity in PDE neurons, suggesting that Rac activity might induce the formation of both lamellipodia-like and filopodia-like structures in vivo. Possibly, Rac activity is required to produce these structures in growth cones and is normally precisely controlled by upstream regulators such as UNC-73 Trio.

Loss of rac function caused both axon pathfinding defects and ectopic axon formation. Constitutive rac activity also caused both axon pathfinding defects and ectopic axon formation, suggesting that Rac activity is required for both axon initiation and growth, as well as for inhibition of superfluous axons and branches. Alternatively, constitutive activation of one rac might trigger a negative regulatory system that attenuates all rac signaling. If racs are required for both axon initiation and axon suppression, the mechanism of ectopic axon formation caused by rac loss-of-function and rac(G12V) constitutive activation might be distinct: rac loss-of-function ectopic axons might be due to a failure to prune spurious axon initiation, whereas Rac(G12V)activity might induce ectopic axons via uncontrolled axon initiation.

Rac(G12V) molecules might persistently signal to their downstream effectors leading to the ectopic formation of cellular structures that are normally precisely controlled during axon pathfinding. We found that UNC-115 is required for all of the morphogenetic effects of RAC-2(G12V), indicating that UNC-115 acts downstream of RAC-2 to mediate these events and that UNC-115 is normally involved in the formation of Rac-induced membrane extensions.

unc-115 did not suppress mig-2(G12V) orced-10(G12V), consistent with the idea that UNC-115 acts with RAC-2 specifically. However, unc-115 might act in the ced-10pathway in parallel to another gene with overlapping function. Possibly,unc-115 does not suppress ced-10(G12V) becauseced-10(G12V) can exert its influence through a downstream effector gene that is redundant with unc-115. Inherent in both models is the existence of other molecules, possibly other actin-binding proteins, that act downstream of MIG-2 and CED-10 in parallel to UNC-115. unc-115 is the only member of the unc-115/abLIM family present in the C. elegans genome (The C. elegans Genome Sequencing Consortium, 1998), sounc-115 redundancy will be at the functional level and not a result of homologous genes (as observed with rac redundancy).

Our results show that UNC-115 is required for the formation of plasma membrane extensions that resemble lamellipodia and filopodia, actin-based structures that normally regulate cell shape. We show that the modeled UNC-115 VHD structure contains a `positive patch' found in other actin-binding VHDs(Vader et al., 2002) and that the UNC-115 VHD binds to actin filaments in vitro, indicating that UNC-115 is an actin-binding protein. UNC-115 might control axon pathfinding by directly interacting with the actin cytoskeleton of growth cones.

The loss-of-function and epistasis suppression experiments described here implicate UNC-115 as a downstream effector of Rac signaling in axon pathfinding. In response to a signal (possibly an extracellular guidance signal), UNC-73 Trio might act on all three Racs, which then influence the actin cytoskeleton to achieve morphogenetic change underlying growth cone outgrowth and steering. UNC-115 might respond to RAC-2 and possibly CED-10 by binding to and modulating the actin cytoskeleton of the growth cone. Other actin binding proteins might act redundantly with UNC-115 to mediate cytoskeletal change in response to CED-10 and MIG-2 signals. Furthermore,UNC-115 appears to act downstream of Rac signaling specifically in axon pathfinding, indicating that Racs might use different downstream effectors to mediate different morphogenetic events.

We thank G. Lushington of the KU Molecular Modeling and Graphics Laboratory for UNC-115 VHD modeling; V. Corbin, M. Herman, L. Timmons and two anonymous reviewers for critical reading of this manuscript; P. Reddien for communicating unpublished results; M. Welch for advice about actin sedimentation assays; A. Fire for gfp vectors; and J. Culotti and C. Kenyon for nematode strains. Some nematode strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources (NCRR). We give special thanks to C. Bargmann for support and many stimulating discussions. This work was supported by an NIH grant (NS4095) to E. A. L.