The Abelson tyrosine kinase, the Trio GEF and Enabled interact with the Netrin receptor Frazzled in Drosophila

Extracellular guidance cues, their receptors on the growth cone surface,and intracellular effectors function together to regulate directional axon extension (Dent and Gertler,2003; Guan and Rao,2003; Huber et al.,2003; Lee and Van Vactor,2003; Luo, 2002). Genetic screens for mutants defective in axon pathfinding at the midline in the Drosophila embryo have identified many of these evolutionarily conserved molecules, and suggest that growth cones respond to a balance of extracellular matrix chemoattractants (e.g. Netrins) and chemorepellents (e.g. Slit) (Araujo and Tear, 2003). frazzled (fra) encodes an attractive Netrin receptor related to mammalian Deleted in Colorectal Cancer (DCC) and C. elegans UNC-40(Chan et al., 1996; Keino-Masu et al., 1996; Kolodziej et al., 1996). Mutations that remove both of the closely related Netrin genes, or mutations in fra, result in thin or missing commissural axon bundles,reflecting a decrease in growth cone attraction to the central nervous system(CNS) midline (Harris et al.,1996; Kolodziej et al.,1996; Mitchell et al.,1996). Conversely, members of the Roundabout (Robo) family of Slit receptors mediate repulsion, and mutations in slit or in robo receptors cause CNS growth cones to cross the midline inappropriately (Battye et al.,1999; Kidd et al.,1999; Kidd et al.,1998; Rajagopalan et al.,2000; Rothberg et al.,1988; Seeger et al.,1993; Simpson et al.,2000a; Simpson et al.,2000b).

An emerging theme in axon guidance is that growth cone receptors recruit cytoplasmic effectors to modulate reorganization of the actin cytoskeleton(Huber et al., 2003; Patel and Van Vactor, 2002). Relatively little is known about how Frazzled and its homologs DCC and UNC-40 signal to the cytoskeleton during axon guidance and outgrowth. A screen in C. elegans for genetic suppressors of an UNC-40 gain-of-function phenotype identified molecules that may function with UNC-40 and Netrin/UNC-6 to regulate actin dynamics (Gitai et al.,2003). These include the actin-binding protein AbLIM/UNC-115,Enabled (Ena)/UNC-34, and the Rho-family guanosine triphosphatase (GTPase)Rac/CED-10. AbLIM/UNC-115 behaves genetically as an effector of signaling by the Rac-2 GTPase (Struckhoff and Lundquist, 2003). The vertebrate orthologs of Ena/UNC-34, Mena,vasodilator-stimulated protein (VASP) and Ena/VASP-like (EVL), antagonize F-actin capping and allow F-actin filament elongation(Bear et al., 2002; Gitai et al., 2003). Netrin stimulation of cultured mouse neurons results in Ena/VASP-dependent filopodia formation and Mena phosphorylation at a protein kinase A regulatory site(Lebrand et al., 2004). In cultured vertebrate cells, the adaptor Nck1 and the GTPases Cdc42 and Rac1 affect Netrin- and DCC-dependent neurite outgrowth, cell spreading and filopodia extension (Li et al.,2002a; Li et al.,2002b; Shekarabi and Kennedy,2002). Nck1 binds DCC in vitro, and can regulate actin nucleation in concert with Rac through WAVE1, an Arp2/3 complex activator; however, it is not known whether the WAVE complex is activated in response to Netrin-DCC signaling (Eden et al., 2002; Li et al., 2002a). Similarly,although Netrin stimulation of DCC-expressing non-neuronal cells leads to activation of Cdc42 and Rac1, the mechanisms by which DCC regulates small GTPase activity have not been elucidated(Li et al., 2002b; Shekarabi and Kennedy,2002).

Other pathways from DCC to the F-actin cytoskeleton are likely to involve cytoplasmic tyrosine kinases. DCC interacts with focal adhesion kinase (FAK),Src and Fyn, and DCC is tyrosine phosphorylated in cells expressing increased levels of these kinases or upon Netrin stimulation; furthermore,phosphorylation of DCC is required for attractive axon turning in cultured neurons and Rac1 activation in non-neuronal cells(Li et al., 2004; Liu et al., 2004; Meriane et al., 2004; Ren et al., 2004). Tyrosine phosphorylation of UNC-40 has also been observed, and genetic interactions indicate that UNC-40 signaling is regulated by the receptor protein tyrosine phosphatase (RPTP) CLR-1 (Chang et al.,2004; Tong et al.,2001).

In Drosophila, signaling by Fra and the Netrins is even less understood. Genetic interactions with fra suggest that Gef64C,weniger, Arf6-Gef/Schizo, Myosin Light Chain Kinase(Stretchin-Mlck – FlyBase) and the G-protein Gαq (Gαq49B – FlyBase)promote commissure formation (Bashaw et al., 2001; Hummel et al.,1999a; Hummel et al.,1999b; Kim et al.,2002; Onel et al.,2004; Ratnaparkhi et al.,2002). However, none of the molecules encoded by these genes nor any others have been linked biochemically to Fra signaling.

The Drosophila Abelson cytoplasmic tyrosine kinase (Abl), the Trio Rac/Rho guanosine-exchange factor (GEF) and Ena are expressed in the nervous system and interact genetically and/or biochemically with receptors known to regulate nervous system development(Awasaki et al., 2000; Bashaw et al., 2000; Bateman et al., 2000; Crowner et al., 2003; Gertler et al., 1989; Gertler et al., 1995; Liebl et al., 2003; Wills et al., 1999). These molecules and their homologs in other organisms regulate cytoskeletal dynamics during diverse developmental processes(Bateman and Van Vactor, 2001; Hakeda-Suzuki et al., 2002; Kwiatkowski et al., 2003; Lanier and Gertler, 2000; Moresco and Koleske, 2003; Van Etten, 1999; Woodring et al., 2003). In cultured cells, these molecules regulate cell migration, neurite extension and leading edge actin dynamics (Bateman and Van Vactor, 2001; Estrach et al., 2002; Kwiatkowski et al.,2003; Moresco and Koleske,2003).

In this study, we expand the understanding of the signaling networks in which Abl, Trio and Ena function by uncovering genetic and biochemical interactions between these molecules and the Netrin receptor Fra. Our results indicate that Abl, Trio and Ena probably function as effectors of Fra signaling in commissural axons, in addition to roles downstream of other growth cone receptors. Furthermore, our observations suggest potential mechanisms by which Fra and other receptors might coordinate actin cytoskeletal dynamics through these molecules.

The following alleles/chromosomes were used: fra⁴; Df(2R)vg135,nompA^vg135 (Df(2R)vg135 is a chromosomal deficiency that removes fra); Df(1)NP5 (removes NetA and NetB); ena^GC10; ena^GC5; ena²¹⁰; ena²³; Abl¹; Abl⁴; Df(3L)st-j7 (removes Abl); trio^M89; Df(3L)FpaI (removes trio); trio^P0368/10; trio^IMP159.4 (an imprecise excision allele generated by mobilizing the P-element in trio^P0368/10); trio^M89,Abl¹; Df(3L)FpaI,Abl⁴; trio^IMP159.4,Abl¹; and UAS-Robo-Fra-Myc (kindly provided by Greg Bashaw). fra⁴,ena^GC10 recombinant chromosomes were generated by meiotic recombination and isolated on the basis of their failure to complement both ena^GC8 and Df(2R)vg135.

All flies were maintained in standard cornmeal-yeast medium at room temperature. Embryos were fixed in 4% paraformaldehyde/1 ×PBS, and the CNS was visualized using mAb BP102 (1:20, Developmental Studies Hybridoma Bank, University of Iowa), anti-β-galactosidase (1:500, Promega), goat anti-mouse-HRP (1:500, Jackson), and standard immunohistochemical procedures(Patel et al., 1987). All alleles were maintained over lacZ-expressing balancers to distinguish the genotype of embryos. Stage 14-16 embryos were filleted and scored at 400× magnification.

pMET Abl-Myc, pMET Trio-Myc, pMET Trio^ΔSPR-Myc, pMET Fra-Myc, pMET Fra-HA, pMET Fra^ΔCYTO-HA [deleted for amino acids P1123-C1375 (GenBank Accession Number U71001)], pBSK Abl-Myc, pBKS Trio-Myc, pBKS Trio^ΔSPR-Myc [deleted for amino acids L285-D1199 (GenBank Accession Number AF216663)], pBSK Ena-Myc,pGEX2T-Fra_CYTO [amino acids C1098-C1375 (GenBank Accession Number U71001)], pGEX2T-AblSH3 [amino acids E202-K268 (GenBank Accession Number AH001049)], and pGEX2T-TrioSH3 [amino acids E1177-L1840, deleted for GEF1(A1281-P1596) (GenBank Accession Number AF216663)] were all constructed using standard molecular techniques; details are available upon request. pPAC Ena was provided by A. Comer (Comer et al.,1998). pMET Fra constructs were generated using the short isoform that rescues fra mutant phenotypes(Kolodziej et al., 1996). Myc and HA tags were added C terminally.

GST and GST-Fra_cyto were generated in E. coli (BL21),as described in Amersham Pharmacia's Gene Fusion System Guide. GST pulldowns of in vitro-translated proteins were performed essentially as described(Bashaw et al., 2000), except that non-radiolabeled, epitope-tagged proteins were generated in vitro using TnT T7-coupled rabbit reticulocyte lysate system (Promega), and Abl, Trio and Ena constructs cloned into pBluescript (Stratagene). An aliquot from each reaction (15-25 μl) was added to ∼10 μg fusion protein bound to beads suspended in 200 μl binding buffer. Binding was overnight, and, after washing, ∼20% of total protein was separated by SDS-PAGE. For GST pulldowns from S2 cell extracts, 2×10⁷ S2 cells were transiently transfected with pMET Abl-Myc, pMET Trio-Myc, or pPAC Ena with CellFectin Reagent (Invitrogen). Twenty-four hours after induction, cells were lysed in 1 ml IP buffer (Comer et al.,1998), and lysates were pre-cleared with 100 μl of Glutathione Sepharose4B beads prior to GST pulldowns.

For co-immunoprecipitations, 2×10⁷ S2 cells were transiently transfected with the relevant constructs, and 24 hours after induction, cells were rinsed once in 1×PBS then lysed in 50 mM Tris (pH 8), 100 mM NaCl, 1 mM MgCl₂, 1% NP-40, 10 mM NaF, 2 mM Na₃VO₄, and 5 μg/ml each of Aprotinin and Leupeptin(Roche). Cell extracts were cleared by centrifugation, lysates were pre-cleared with 40 μl Protein G Sepharose beads (Sigma) for 30 minutes,and protein complexes were immunoprecipitated with 1 μg rabbit anti-Myc(Santa Cruz), anti-Ena [5G2, Developmental Studies Hybridoma Bank(Bashaw et al., 2000)] diluted 1:20, or anti-HA (HA.11, Covance) diluted 1:150, for 60 minutes. Immune complexes were recovered on 40 μl Protein G beads for 60 minutes, washed three to four times in lysis buffer containing 1 μg/ml Aprotinin and Leupeptin, and boiled in 6× sample buffer.

For Trio and Fra phosphorylation experiments, 2×10⁷ S2 cells were transiently transfected with pMET Trio-Myc, pMET Fra-Myc, pMET Abl,and/or empty pMET vector (to control for transfection efficiency in experiments receiving less than 5 μg pMET Abl). For pervanadate treatment,cells [in Schneider's Media (BRL/Invitrogen), plus 10% FBS] were treated with 2 mM Na₃VO₄ and 3 mM H₂O₂ for 30 minutes at room temperature. In all experiments, cells were rinsed once in PBS and then lysed in 50 mM Tris (pH 8), 300 mM NaCl, 5 mM EDTA, 1% Triton X-100,10 mM NaF, 2 mM Na₃VO₄, and 5 μg/ml each of Aprotinin and Leupeptin. After immunoprecipitation (IP), beads were washed four to six times in lysis buffer containing 1 μg/ml Aprotinin and Leupeptin, and boiled in 2 × Sample Buffer.

Phosphotyrosine was detected using 4G10 mouse anti-phosphotyrosine(Upstate) at 1:10,000 and goat anti-mouse-HRP (Jackson) at 1:10,000 in low-salt TBST (25 mM NaCl)/5% BSA. Mouse anti-Myc (Roche), rabbit anti-Myc(Santa Cruz), mouse anti-HA, and rabbit anti-Abl (kindly provided by A. Comer)were used at a dilution of 1:2000 in 5% milk/TBST. Mouse anti-Ena 5G2 was used at 1:200 in 5% milk/TBST. Proteins were visualized using ECL (for phosphotyrosine detection), ECL PLUS (Amersham Pharmacia, for protein-protein interactions), or NBT/BCIP detection (for loading controls in co-IP and phosphorylation experiments). Prior to the re-probing of co-IP and phosphotyrosine westerns, blots were stripped in 50 mM Tris-HCl (pH 6.8), 2%SDS and 100 mM β-mercaptoethanol overnight at 65°C.

Noting that the trio,Abl double mutant CNS phenotype is qualitatively similar, but much more severe, than the phenotype of fra mutant embryos, and that Fra, Trio and Abl localize to CNS axons,we hypothesized that these molecules collaboratively regulate commissure formation (Fig. 1A-F; Tables 1, 2, 4)(Awasaki et al., 2000; Gertler et al., 1989; Kolodziej et al., 1996; Liebl et al., 2000). Because dosage-sensitive genetic interactions often indicate that gene products function in the same biological process, we asked whether mutations in Abl or trio dominantly modify the fra CNS phenotype.

In homozygous fra (fra⁴/Df(2R)vg135) mutant embryos, 21% of segments had defective commissures. As in a previous study(Kolodziej et al., 1996), we found that the majority of defective segments in fra embryos (13% in this case) had thin or missing commissures, and that the posterior commissure was most often affected (Fig. 1F, Table 1). In 8%of segments, an approximately wild-type number of axons crossed the midline,but commissures were disorganized (examples of errors are indicated in Fig. 1 and are described in the Table 1 footnotes).

In homozygous fra, heterozygous Abl(fra⁴/Df(2R)vg135; Abl⁴/+) embryos, the percentage of segments with defective commissures increased to 30%, and this increase was due solely to an increase in the number of thin or missing commissures (24%, Fig. 1G, Table 1). The independently generated Abl¹ allele also dominantly enhanced the fra phenotype, demonstrating that lesions in Abl were responsible for the genetic interaction with fra, and not accessory mutations on any of the chromosomes tested(Table 1).

Heterozygosity for trio also enhanced the CNS phenotype in fra mutant embryos. For example, in fra⁴/Df(2R)vg135;trio^IMP159.4/+ embryos, 53% of segments had defective commissures (Fig. 1H, Table 2). Forty-two percent of segments in fra⁴/Df(2R)vg135;trio^IMP159.4/+embryos had thin or missing commissures (versus 13% in fra⁴/Df(2R)vg135 animals), and 11% of segments had commissural pathfinding errors (versus 8% in fra⁴/Df(2R)vg135 embryos)(Fig. 1H, Table 2). Milder dominant enhancement of the fra phenotype was observed for a number of other trio alleles, including the deficiency Df(3L)FpaI, the hypomorphic P-element insertion allele trio^P0368/10, and the trio^M89 allele (which encodes a point mutation in the GEF1 domain of Trio) originally identified as a dominant enhancer of the Abl semilethality phenotype(Liebl et al., 2000)(Table 2). It is not clear why trio^IMP159.4 (an imprecise excision allele generated by mobilizing the P element on the trio^P0368/10 chromosome)enhances the fra phenotype so strongly. As the deficiency Df(3L)FpaI (which completely removes the trio gene) behaves similarly to the other trio alleles we tested, it is likely that trio^IMP159.4 is not a null or hypomorphic allele, but rather encodes a Trio protein with unusual properties. It is also not clear why fra⁴ homozygotes have more disrupted commissures than fra⁴/Df(2R)vg135 animals (this fra allele has not been characterized), although fra⁴ homozygous animals are not immunoreactive with the polyclonal anti-fra serum generated by Kolodziej et al. (Kolodziej et al.,1996).

Although in most fra mutant combinations tested the posterior commissure (PC) was affected more often than the anterior commissure (AC),heterozygosity for Abl or trio increased the frequency of defects in both commissures. For example, in fra⁴/fra⁴ embryos 20% (n=212) of PCs scored were thin or missing, versus 31% (n=215) in fra⁴/fra⁴;Abl¹/+ mutants and 39%(n=262) in fra⁴/fra⁴;trio^M89/+ embryos. Similarly, 5% (n=212) of ACs scored were thin or missing in fra⁴/fra⁴ embryos, versus 14% (n=215)in the fra⁴/fra⁴;Abl¹/+ background and 18% (n=262) in fra⁴/fra⁴;trio^M89/+ embryos. Thus,although mutations in fra seem to affect posterior commissural axons preferentially (Kolodziej et al.,1996), the genetic interactions of fra with Abland trio are consistent with a significant role for all of these genes in anterior commissure formation as well.

We next asked whether heterozygosity for Abl or triomodifies the CNS phenotype in embryos mutant for genes encoding the Frazzled ligands Netrin A and Netrin B. NetA and NetB are both removed by a deficiency on the X chromosome, Df(1)NP5(Mitchell et al., 1996). Like fra mutant embryos, Netrin mutant embryos have fewer commissural axons, with those in the posterior commissure being most affected(Mitchell et al., 1996). As observed for fra, both Abl and trio dominantly enhanced the Netrin deficiency phenotypes and increased the frequency of defects in both commissures (Fig. 1P-R, and data not shown). In Df(1)NP5/Y;Abl⁴/+ embryos, 33% of segments had thin or missing commissures, versus only 24% in Df(1)NP5/Y hemizygotes; in Df(1)NP5/Y;trio^IMP159.4/+embryos, 56% of segments had thin or missing commissures (Tables 1, 2; Fig. 1P-R). Genetic interactions of Abl and trio with the Netrin genes further support the idea that Abl and trio are required to attract growth cones to the CNS midline.

Although mutations in Abl and trio dominantly enhanced the loss-of-commissure phenotype in fra and Netrin mutants,we did not observe other dose-sensitive interactions between fra, Abland trio. For example, mutations in fra did not reciprocally enhance the loss-of-commissure phenotype in Abl or triomutants, although the percentage of segments with axon pathfinding errors increased when one dose of fra was removed in the Abl mutant background (Tables 1, 2; Fig. 1I,J). We also did not observe transheterozygous interactions between Abl, trio and fra in single, double and triple transheterozygous mutant combinations. In all of these cases, no more than 5% of segments had disrupted commissures (Table 3, Fig. 1K). Furthermore, even in embryos that were homozygous mutant for one of these three genes,heterozygosity for the two remaining genes did not lead to additive or synergistic increases in commissure defects(Table 3). Because in other experiments, Abl and trio behaved genetically as fra effectors (see below), the inability of fra to dominantly enhance Abl or trio loss-of-commissure defects,and the lack of transheterozygous interactions between fra, Abl and trio, may be due to the presence of maternally-contributed Abl and trio in the embryo(Bennett and Hoffmann, 1992; Liebl et al., 2000; Wadsworth et al., 1985). Another possibility is that heterozygosity for fra simply does not reduce the effective dose enough to enhance Abl or trioloss-of-commissure phenotypes.

As in trio,Abl double mutant embryos, fra;Abl and fra;trio double mutant embryos had severe CNS phenotypes in which the majority of segments had thin or missing commissures. In fra⁴/Df(2R)vg135;Abl⁴/Abl¹ embryos,85% of segments had thin or missing commissures(Fig. 1L,M; Table 1). In fra⁴/Df(2R)vg135;trio^IMP159.4/trio^M89embryos, 66% of segments had thin/missing commissures(Fig. 1N,O; Table 2). In these animals, as in homozygous fra, heterozygous Abl or triomutants, the posterior commissure was affected more often than the anterior commissure. For example, in fra⁴/Df(2R)vg135;Abl⁴/Abl¹ embryos,79% of posterior commissures were thin/missing, versus only 49% of anterior commissures (n=166 segments); in fra⁴/Df(2R)vg135;trio^IMP159.4/trio^M89 embryos, 62% of posterior commissures were thin/missing, versus only 33% of anterior commissures(n=172 segments).

Additionally, although there were occasional breaks (>75% of axons missing) in longitudinal connectives in fra homozygotes [10%(n=519) in fra⁴/Df(2R)vg135 embryos], in fra;Abl and fra;trio double mutants this type of defect did not increase considerably. For example, only 11% (n=387) of connectives had breaks in fra⁴/Df(2R)vg135;Abl⁴/Abl¹ embryos,and 16% (n=368) in fra⁴/Df(2R)vg135;trio^IMP159.4/trio^M89embryos. However, analysis of subsets of longitudinally-projecting axons in stage 17 embryos using the monoclonal antibody 1D4 (anti-Fasciclin II)revealed a significant disorganization of these pathways, especially in fra;Abl double mutants (see Fig. S1 and Table S1 in the supplementary material). In fra;Abl and fra;trio double mutants,Fas2-positive longitudinal pathways wandered medially or laterally, often seeming to intertwine so that individual bundles were indistinguishable. In addition, although breaks in all three longitudinal bundles between segments were rare, often one to two (usually lateral) 1D4-positive bundles were discontinuous between segments. Similar defects were observed at a lower frequency in individual fra, Netrin, Abl and trio mutants(see Fig. S1 and Table S1 in the supplementary material). Thus, although BP102 immunohistochemistry did not reveal consistent defects in longitudinal pathways, the disorganization of Fas2-positive axon bundles suggests that Fra,Abl and Trio function during axon pathfinding in longitudinal pathways in addition to their roles during commissure formation.

In the CNS, ena interacts genetically with the repulsive receptor robo, leading to inappropriate crossing of the midline by longitudinal axons (Bashaw et al.,2000). In Abl, trio and fra mutant combinations,numerous axons fail to cross the midline(Liebl et al., 2000) (Tables 1, 2). We explored this apparently antagonistic relationship further by analyzing genetic interactions among trio, Abl, fra, the Netrin genes and ena in the CNS.

Mutations in ena dominantly reduced the severity of the CNS phenotype in trio,Abl mutants. For example, in enaheterozygous, trio,Abl homozygous(ena^GC10/+;trio^M89,Abl¹/Df(3L)FpaI,Abl⁴)embryos, only 78% of segments had defective commissures, versus 100% in the trio,Abl(trio^M89,Abl¹/Df(3L)FpaI,Abl⁴)double mutant (Table 4; and compare Fig. 1D,E with 1T). Overall, there was a 20% reduction in the number of segments with thin or missing commissures (Table 4). However, analysis of individual commissures revealed a dramatic increase in the number of axons which crossed the midline when compared with the trio,Abl double mutant, especially in the anterior commissure. For example, in ena heterozygous, trio,Abl homozygous(ena^GC10/+;trio^M89,Abl¹/Df(3L)FpaI,Abl⁴)embryos, only 24% of anterior commissures and 59% of posterior commissures were thin or missing (n=206 segments), compared with 64% of anterior commissures and 84% of posterior commissures in trio,Abl embryos(n=160 segments) (Fig. 1D,E,T). In another mutant combination, the increase in the number of axons crossing the midline was even more striking. In Df(3L)FpaI,Abl⁴/trio^IMP159.4,Abl¹embryos, 65% of segments had thin or missing commissures, whereas in ena^GC5/+;Df(3L)FpaI,Abl⁴/trio^IMP159.4,Abl¹embryos only 9% of segment commissures were thin or missing(Table 4). Removing one dose of ena in the trio,Abl homozygous mutant background also decreased the number of breaks in longitudinal connectives(Fig. 1D,E,T; and see Table S1 in the supplementary material).

Mutations in ena also dominantly suppressed CNS defects in fra and Netrin mutants. For example, in Df(1)NP5/Y;ena^GC10/+ embryos, only 7% of segments had thin/missing commissures, versus 24% in Df(1)NP5/Y embryos(Fig. 1P,S, and Table 4). In fra⁴,ena^GC10/Df(2R)vg135 embryos, 10% of segments had thin or missing commissures, versus 13% in fra⁴/Df(2R)vg135 animals(Table 4). Heterozygosity for ena also suppressed enhancement of the fra CNS phenotype by Abl or trio. For example, in fra⁴,ena^GC10/fra⁴;Abl⁴/+embryos 14% of segments had thin/missing commissures, versus 33% in fra⁴/fra⁴;Abl⁴/+ mutants, and 23% in fra⁴/fra⁴ embryos. Similarly, in fra⁴/fra⁴;trio^IMP159.4/+ embryos,80% of segments had thin or missing commissures, compared with only 15% in fra⁴,ena^GC10/fra⁴;trio^IMP159.4/+embryos (Table 4).

The genetic interactions described support roles for Abl, Trio and Ena during commissure formation, but these types of interactions (dominant enhancement/suppression and synergistic double mutant genetic interactions)are usually interpreted to mean that gene products function in parallel pathways; it is likely, for example, that Abl and Trio are instructed by at least one other receptor that positively regulates commissure formation. To test genetically whether Abl, Trio and/or Ena might function as Fra effector(s) in vivo, we took advantage of the fact that neuronal expression of a chimeric receptor composed of the extracellular and transmembrane domains of the repulsive Robo receptor and the intracellular domain of Fra causes CNS axons to inappropriately sense the midline-secreted repellent Slit as an attractant and cross the midline boundary inappropriately(Bashaw and Goodman, 1999). If Abl, Trio and/or Ena function as effectors of Fra signaling in axons that cross the midline, then reducing the dose of these molecules genetically would be expected to reduce the severity of the chimeric Robo-Fra receptor phenotype. We chose this strategy because, in the Drosophila embryo,overexpressing full-length Fra in CNS neurons does not cause robust ectopic midline crossing by CNS axons (D.J.F. and P.A.K., unpublished)(Kim et al., 2002).

In wild-type stage 17 embryos, Fas2-positive axons project longitudinally in three bundles on either side of the midline, but these axons never cross the midline boundary (Fig. 2A). In embryos expressing UAS-Robo-Fra in all neurons, numerous Fas2-positive axon bundles crossed the midline inappropriately(Table 5, Fig. 2B). Reducing the gene dose of Abl in embryos expressing the chimeric receptor led to a moderate reduction in the number of these ectopic crossovers(Table 5). Similarly, in three out of four alleles tested, heterozygosity for trio also reduced the severity of the Robo-Fra phenotype (Table 5). In these experiments, the deficiency Df(3L)FpaIsuppressed the Robo-Fra receptor phenotype most strongly, and not the imprecise excision allele trio^IMP159.4, which acted as the strongest dominant enhancer of the fra loss-of-function phenotype(Table 2). It is possible that other, unidentified genes removed by the FpaI deficiency also function as Fra effectors, that the other trio alleles encode Trio proteins that retain partial function downstream of Fra signaling, or that trio^IMP159.4 disrupts signaling by other receptor(s) that mediate commissure formation more strongly than this allele interferes with Fra signaling. We also discovered that if the Df(3L)FpaI chromosome was contributed to progeny by the male parent, rather than the female, genetic suppression of the Robo-Fra phenotype was less severe(Table 5), suggesting that maternal contribution of trio (or another gene removed by this deficiency) plays a role.

Interestingly, although heterozygosity for Abl⁴ [the Abl point mutant allele thought to be protein-null or nearly null(Bennett and Hoffmann, 1992)]led to a negligible reduction in ectopic crossovers on its own (0.84 crossovers/segment), heterozygosity for this allele of Abl and three different trio mutations led to a moderately synergistic reduction in the number of inappropriate crossovers. For example, in ELAV-GAL4;UAS-Robo-Fra; +,Abl⁴/trio^P0368/10/+ embryos,only 0.47 axon bundles per segment crossed the midline inappropriately,compared with 0.61 crossovers/segment in ELAV-GAL4; UAS-Robo-Fra; trio^P0368/10/+ embryos(Table 5, Fig. 2C). These data are consistent with positive roles for both Abl and Trio as effectors of Fra signaling in axons that cross the CNS midline.

Heterozygosity for three different alleles of ena also reduced the severity of the Robo-Fra phenotype (Table 5). Furthermore, reducing the gene dose of both Abl and ena also led to a synergistic reduction in the number of ectopic midline crossovers by Fas2-positive axons, similar to the genetic interaction between Abl and trio in embryos expressing Robo-Fra. For example, in ELAV-GAL4; UAS-Robo-Fra/ena²³;Abl⁴/+ embryos, there were only 0.23 crossovers per segment,compared with 0.47 crossovers/segment in Robo-Fra-expressing embryos heterozygous for ena²³ only(Table 5, Fig. 2D). While these data were initially surprising because heterozygosity for ena led to an increase in the number of axons that crossed the midline in fra, Netrin,Abl and trio loss-of-function mutants(Table 4), they are consistent with the idea that Ena may be functioning positively as an effector of signaling via the Fra cytoplasmic domain, similar to orthologs of Ena in other organisms (see Discussion) (see also Gitai et al., 2003; Lebrand et al.,2004).

To test whether Fra could physically interact with Abl, Trio or Ena, we first asked whether glutathione-S-transferase (GST) fusions with the cytoplasmic domain of Fra (generated in E. coli) could bind in vitro translated Abl, Trio or Ena. In these experiments, Abl and Trio, but not Ena,specifically bound GST-Fra_CYTO, but not GST or glutathione beads alone (Fig. 3A and data not shown), indicating that the cytoplasmic domain of Fra can interact directly with Abl and Trio. In parallel experiments, we also found that GST-Fra could interact with Abl-Myc and Trio-Myc in extracts from DrosophilaSchneider-2 (S2) cells that had been engineered to express each protein (not shown).

In addition, in vitro translated Abl-Myc bound specifically to GST-TrioSH3(Fig. 3B), and Trio-Myc bound specifically to GST-AblSH3 (Fig. 3C), indicating that Abl and Trio can interact directly as well via their SH3 domains. GST-TrioSH3 includes linker sequences on either side of the SH3 domain. Deleting most of the SH3 domain and C-terminal linker sequences in GST-TrioSH3 (L1624-L1840) abolishes Abl-Myc binding (not shown),suggesting that the SH3 domain mediates the interaction with Abl. In all of the GST-pulldown experiments, the Trio constructs used were deleted for the spectrin-like repeats in order to maximize expression level and stability(full-length, the molecular mass of wild-type Trio is greater than 225 kDa). Thus, at least in these in vitro assays, the spectrin-like repeats are not required for the interaction of Trio with either the cytoplasmic domain of Fra or the SH3 domain of Abl.

Next, we investigated whether full-length Fra could interact with full-length Abl, Trio and/or Ena in S2 cells. Hemagglutinin (HA)-tagged Fra was transiently co-expressed with either Abl-Myc, Trio-Myc or Ena, and extracts were subjected to immunoprecipitation. Fra-HA specifically co-immunoprecipitated in the presence of Abl-Myc or Trio-Myc, but not in their absence (Fig. 3D,E, lanes 1 and 2). This interaction is mediated by the cytoplasmic domain of Fra, as a Fra protein deleted for this domain did not co-immunoprecipitate with either Abl or Trio (Fig. 3D,E, lane 3). We did not observe a similar association between Ena and Fra-HA in S2 cells,using either anti-Ena or anti-HA antibodies for co-immunoprecipitation (data not shown).

The genetic and physical interactions that we observed among Fra, Trio and Abl raised the possibility that Fra and Trio might be substrates for the Abl tyrosine kinase. To determine whether or not phosphotyrosine could be detected on Fra or Trio, we transiently expressed full-length, epitope-tagged versions of Trio and Fra in S2 cells, and treated the cells with pervanadate, a potent phosphotyrosine phosphatase inhibitor that has been used previously to sustain tyrosine phosphorylation of proteins in Drosophila cells(Fashena and Zinn, 1997; Muda et al., 2002). In control S2 cells, Trio tyrosine phosphorylation was not detected, whereas Fra was moderately tyrosine phosphorylated (Fig. 4A,B, lane 1). After a 30-minute pervanadate treatment, both molecules were robustly tyrosine phosphorylated, indicating that Trio and Fra are substrates of tyrosine kinases and phosphatases expressed endogenously in S2 cells (Fig. 4A,B, lane 2).

Next, we investigated whether elevating Abl levels in S2 cells would increase tyrosine phosphorylation of Trio or Fra. Phosphotyrosine on Trio increased dramatically when Abl was co-expressed(Fig. 4C, lane 2). Fra was also tyrosine phosphorylated at a higher level when Abl was co-expressed, although not quite as robustly as Trio (Fig. 4D, lanes 2 and 3). We observed a similar elevation of tyrosine phosphorylation of Ena, a known Abl substrate, in S2 cells treated with pervanadate or cotransfected with Abl (not shown). These results indicate that Abl either phosphorylates Fra and Trio directly, or indirectly regulates Fra and Trio tyrosine phosphorylation.

Recent investigations into the molecular mechanisms of axon guidance have focused on identifying which cytoplasmic molecules cooperate with which growth cone receptors to regulate actin cytoskeletal dynamics and growth cone motility (Huber et al., 2003). In this study, we present evidence that Abl, Trio and Ena function together with the Netrin receptor Fra to regulate chemoattraction to the Drosophila embryonic CNS midline.

We found that mutations in Abl and trio dominantly enhance the CNS phenotype in fra and Netrin mutant embryos,and that fra;Abl and fra;trio double mutants have a severe CNS phenotype in which a majority of commissures are thin or missing, similar to the trio,Abl double mutant phenotype. Mutations in Abland trio reduce the number of axons that cross the midline inappropriately in embryos expressing the chimeric Robo-Fra receptor. Abl and Trio interact physically with the cytoplasmic domain of Fra, and increasing Abl kinase expression in cells increases tyrosine phosphorylation of Fra and Trio. Interpreting these data together, we conclude (1) that Abl, Trio and Fra function together during commissure formation, (2) that the severe double mutant phenotypes reflect the disruption of multiple signaling pathways or networks in the growth cones of commissural axons (i.e. Abl and Trio function downstream of at least one other receptor that positively regulates commissure formation), and (3) that the lack of other dose-sensitive interactions between fra, Abl and trio is a result of redundancy (other receptors/effectors mediating commissure formation), or the presence of maternally contributed proteins.

The interactions of Abl with Fra are intriguing, as they suggest that in Drosophila, as in other organisms, this evolutionarily conserved guidance receptor is regulated by tyrosine phosphorylation, and also that Fra may regulate Abl substrates. Recently, others have demonstrated Netrin-dependent tyrosine phosphorylation of DCC, Netrin/DCC-dependent activation of the tyrosine kinases FAK, Src and Fyn, and the requirement of DCC tyrosine phosphorylation for Netrin-dependent Rac1 activation and growth cone turning (Li et al., 2004; Liu et al., 2004; Meriane et al., 2004; Ren et al., 2004). Interestingly, the tyrosine residue in DCC identified as the principal target of Fyn/Src kinases is not conserved in Drosophila Fra or C. elegans UNC-40, suggesting that the precise mechanisms by which Fra/DCC/UNC-40 signaling is regulated by tyrosine kinases may differ between organisms (Li et al., 2004; Meriane et al., 2004). Tyrosine phosphorylation of UNC-40 has also been observed, and although the kinase(s) responsible has not been identified, genetic interactions suggest that UNC-40 signaling is regulated by the RPTP CLR-1, supporting the idea that regulation of tyrosine phosphorylation is a consequence of UNC-6/Netrin signaling in C. elegans as well(Chang et al., 2004; Tong et al., 2001). In this study, we observed more robust tyrosine phosphorylation of Fra in cells with pervanadate stimulation than with Abl overexpression alone, raising the possibility that additional kinase(s) may function during Fra signaling. Further investigation will be needed to address this issue and to determine how Abl-mediated phosphorylation of Fra modulates commissural growth cone guidance.

Abl is thought to control actin dynamics in part through its ability to regulate other proteins through tyrosine phosphorylation(Lanier and Gertler, 2000; Woodring et al., 2003). Thus,in addition to potential regulation of Fra, Fra may recruit Abl to regulate other Abl substrates. Abl interacts genetically with trio(Liebl et al., 2000), and in this study, we have found that Trio physically interacts with Abl in vitro and that Trio tyrosine phosphorylation increases dramatically with co-expression of Abl. Phosphorylation of Trio may affect its activity, as observed for other GEFs. For example, Abl regulates phosphorylation and Rac-GEF activity of Sos1,and Lck, Fyn, Hck and Syk kinases tyrosine phosphorylate Vav GEF and stimulate its activity (Sini et al.,2004; Turner and Billadeau,2002).

Trio physically interacts with Fra in vitro and in S2 cells, suggesting that Fra can recruit Trio directly. In addition, heterozygosity for trio dominantly modifies the Robo-Fra chimeric receptor phenotype,consistent with a positive role for Trio as a downstream effector of Fra signaling in vivo. As a Rac/Rho GEF, Trio may link Netrin-Fra signaling to the regulation of Rho-family GTPases in commissural axons. Rho-family GTPases have been rigorously studied with regard to their role in the regulation of cytoskeletal dynamics and axon guidance, outgrowth and branching(Dickson, 2001; Luo, 2000). Although positive roles for GTPases in commissure formation in the Drosophila embryo have not been directly demonstrated, trio (in this study) and GEF64C, a Rho GEF (Bashaw et al.,2001), interact genetically with fra leading to the dramatic disruption of commissures. Additionally, expression of constitutively active or dominantly negative isoforms of both Rac and Rho, as well as constitutively active Cdc42, causes axons to cross the CNS midline inappropriately (Fan et al.,2003; Fritz and VanBerkum,2002; Matsuura et al.,2004). Recent studies have implicated Cdc42 and Rac1/CED-10 as effectors of DCC and UNC-40 signaling, but the biochemical mechanisms by which GTPases are regulated have been elusive(Gitai et al., 2003; Li et al., 2002a; Li et al., 2002b; Shekarabi and Kennedy, 2002). Future experiments must determine whether Netrin-Fra signaling modulates the GEF activity of Trio, and how this occurs.

In this study, we found that reducing the genetic dose of enacauses either more or fewer axons to cross the CNS midline, depending on the genetic background, suggesting that the role of Ena in the growth cone is complex. Heterozygosity for ena in embryos expressing the Robo-Fra chimeric receptor reduces the number of axon bundles that inappropriately cross the CNS midline, consistent with a role for Ena as a positive effector of Fra signaling. Ena/UNC-34 has been identified genetically as an effector of DCC/UNC-40 in C. elegans (Gitai et al., 2003). In cultured mouse neurons, Ena/VASP proteins are required for Netrin-DCC-dependent filopodia formation, and Mena is phosphorylated at a PKA regulatory site in response to Netrin stimulation(Lebrand et al., 2004). In migrating fibroblasts, increasing Ena/VASP proteins at the leading edge leads to unstable lamellae and decreased motility; by contrast, increasing Ena/VASP levels at the leading edge in growth cones causes filopodia formation,possibly due to differences in the distribution of actin bundling or branching proteins (Bear et al., 2000; Bear et al., 2002; Lebrand et al., 2004). Although the role of Ena in actin reorganization in Drosophila has not been rigorously studied, Ena localizes to filopodia tips in cultured Drosophila cells, suggesting that the role of Ena in filopodia formation may be conserved (Biyasheva et al., 2004).

We have not observed a direct biochemical interaction between Fra and Ena. However, Abl binds and phosphorylates Ena, and heterozygosity for both Abl and ena further suppresses the Robo-Fra phenotype,suggesting that Fra may recruit Abl to regulate filopodial extension through Ena (Comer et al., 1998; Gertler et al., 1995). Alternatively, Fra may regulate Ena through other molecule(s), and the synergistic suppression of the Robo-Fra phenotype by Abl and ena is a result of the compromise of parallel pathway(s) regulated by Fra. It is important to note that the functional consequences of biochemical interactions between Abl and Ena are not understood(Comer et al., 1998; Grevengoed et al., 2003; Krause et al., 2003). Therefore it will be of particular interest to determine whether Ena is tyrosine phosphorylated in response to Netrin-Fra signaling, and if Ena phosphorylation regulates its activity during filopodial extension.

In addition to suppressing the Robo-Fra chimeric receptor phenotype,mutations in ena also suppress the loss-of-commissure phenotype in fra, Netrin, trio and Abl mutant combinations. In Drosophila (as well as in C. elegans), Ena interacts genetically and biochemically with the repulsive receptor Robo, indicating that Ena may restrict axon crossing at the midline(Bashaw et al., 2000; Yu et al., 2002). Thus, the fact that mutations in ena dominantly suppress fra, Netrin,trio and Abl CNS phenotypes could simply reflect the compromise of a parallel, opposing signaling pathway. Consistent with this idea, some axons that cross the midline in ena heterozygous, trio,Ablhomozygous embryos are Fas2 positive (D.J.F., unpublished), indicating a partial reduction in repulsive signaling. However, ena also dominantly suppresses fra and Netrin commissural pathfinding defects, without causing longitudinal Fas2-positive axons to cross the midline(D.J.F., unpublished). Reductions in Robo signaling therefore may not fully explain the ability of ena to suppress defects in fra, Netrin,Abl and trio mutants.

Based on the fact that mutations in ena suppress a number of Abl mutant phenotypes, it has been proposed that Abl antagonizes Ena function (Grevengoed et al.,2003; Grevengoed et al.,2001; Lanier and Gertler,2000). In Abl mutant embryos, Ena and actin mislocalize during dorsal closure and cellularization, and apical microvilli are abnormally elongated, indicating that Abl regulates the localization of Ena(Grevengoed et al., 2003; Grevengoed et al., 2001). In migrating fibroblasts, increasing Ena/VASP levels at the leading edge results in long, unbranched actin filaments, unstable lamellae, and decreased motility due to increased antagonism of capping protein(Bear et al., 2000; Bear et al., 2002). Interestingly, mutations in the gene encoding Drosophila capping protein β enhance CNS axon pathfinding defects in Abl mutants,including commissure formation (Grevengoed et al., 2003). Therefore, if Fra and/or Abl regulate Ena localization in commissural axons, then in fra, Netrin or Abl mutants, Ena may be mislocalized in the growth cone, leading to inappropriate inhibition of capping protein and excessive F-actin filament elongation. Additionally, reducing regulation of Ena by Fra or Abl may also allow greater Ena regulation by Slit-Robo signaling. In either case, reducing the gene dose of ena in fra, Netrin and trio,Ablmutant embryos would partially relieve these effects, allowing axons to respond more efficiently to other cues and cross the midline, as we observed. Consistent with this idea, Lebrand et al.(Lebrand et al., 2004) found that either increasing or decreasing Ena/VASP proteins at the leading edge impaired the elaboration of growth cone filopodia in response to Netrin-DCC signaling, suggesting that Ena/VASP levels must be tightly regulated in order for the growth cone to respond optimally to extracellular signals.

The role of Abl in the growth cone is also likely to be complex. Our observations implicate Abl as an effector of attractive Fra signaling. In addition, tyrosine phosphorylation of Robo by Abl is thought to negatively regulate repulsive signaling by Robo(Bashaw et al., 2000). Paradoxically though, loss-of-function mutations in Abl, robo and slit interact genetically, resulting in inappropriate axon crossing at the midline, and indicating that Abl may also promote repulsion in longitudinally migrating growth cones(Hsouna et al., 2003; Wills et al., 2002). Obviously, much remains to be understood about the molecular basis for genetic interactions of Abl, particularly how Abl and its various substrates cooperate with different growth cone receptors to yield specific cytoskeletal outputs.

In summary, we have observed genetic and biochemical interactions indicating that Abl, Trio and Ena are integrated into a complex signaling network with Fra and the Netrins during commissure formation. These observations identify another receptor that acts through these effectors, and provide a framework for further investigation of signaling by this key,evolutionarily conserved guidance receptor.

We are grateful to Drs Francis Fogerty, Greg Bashaw, Allen Comer and F. Michael Hoffmann for their kind gifts of fly stocks, antibodies and cDNA. We thank Dr Anthony Harrington and Dr Sung Ok Yoon for technical advice with biochemical experiments. We thank an anonymous reviewer for suggesting an analysis of genetic interactions with the Robo-Fra chimeric receptor. This work was supported by National Science Foundation Grant IBN-0090239 to M.A.S. and E.C.L., and by National Institutes of Health grant NS32839 to M.A.S.