Abi plays an opposing role to Abl in Drosophila axonogenesis and synaptogenesis

Two key cellular processes are needed to establish neuronal connections:axon guidance and synapse formation. These processes involve extracellular cues and intracellular signaling (Dent and Gertler, 2003; Zhou and Cohan,2004). For intracellular signaling, many actin regulators,including proteins functioning in actin nucleation, elongation and monomer binding, are required for axonal connection in the developing central nervous system (CNS) (Sanchez-Soriano et al.,2007). One actin-binding protein, Abl tyrosine kinase, regulates axonal guidance in the developing Drosophila and mammalian CNS(Hoffmann, 1991; Lanier and Gertler, 2000; Moresco and Koleske, 2003). Many actin regulators, including ena, chickadee (also known as profilin) and capulet, also genetically interact with Abl in this process (Gertler et al.,1995; Wills et al.,1999; Wills et al.,2002). Mammalian Abl (known as Abl1) also functions in dendrite maintenance and synaptic plasticity(Moresco and Koleske, 2003; Moresco et al., 2005),although it is unclear how well these functions are conserved in Drosophila.

Accumulated studies have found that Abelson interacting protein (Abi), a target of Abl family tyrosine kinases, forms a pentameric complex with SCAR/WAVE, Sra-1, Kette/Hem and HSPC300 when regulating actin dynamics(Eden et al., 2002; Innocenti et al., 2004; Kunda et al., 2003; Rogers et al., 2003). Moreover, all members of the complex, except Abi, have been reported to be involved in axonogenesis in Drosophila(Bogdan et al., 2004; Bogdan and Klambt, 2003; Hummel et al., 2000; Qurashi et al., 2007; Schenck et al., 2003; Schenck et al., 2004). In addition, other Abi-binding proteins, including Wasp(Bogdan et al., 2005) and Ena(Juang and Hoffmann, 1999; Tani et al., 2003), are involved in CNS development (Gertler et al., 1995; Ben-Yaacov et al.,2001). Genetic and immunohistochemical studies in mammals have also demonstrated that the Abi family proteins play a role in synaptic development (Grove et al.,2004; Proepper et al.,2007). Together, these reports suggest that Abl and Abi have similar roles in axonogenesis and synaptogenesis. However, no study to date has explored whether abi interacts genetically with Ablduring nervous system development.

In this study we used a multidisciplinary approach, including genetics,immunohistochemistry, electron microscopy, electrophysiology and behavioral analysis, to characterize the functional connection between Abl and Abi in the Drosophila nervous system. Our results support a model in which Abl and Abi play opposing roles in regulating Drosophila axonogenesis and synaptogenesis through actin cytoskeleton reorganization.

The wild-type strain used in this study was w¹¹¹⁸. The abi alleles we used were two P-element insertion lines, GE23319(abi^P1) and GE24211 (abi^P2) (GenExel),one deficiency line, Df(3R)JY19 (hereafter referred to as abi^Df), uncovering the abi locus(Breen and Harte, 1991), and a gene-targeting (knockout, KO) line. The KO allele was created by end-out homologous recombination. Briefly, we excised hsp70-white 2.8 kb cDNA from pBS-70w and subcloned it into a filled-in blunting XhoI site(+262 base pairs relative to translation start site) on abi exon 1. The abi genomic DNA plus hsp70-white was subcloned into the NotI site of pP{EndsOut2} to the create targeting plasmid (see Fig. S1A in the supplementary material). pBS-70w and pP{EndsOut2} plasmids were gifts from Dr Jeff Sekelsky (University of North Carolina, Chapel Hill, NC,USA). The fly transformation and targeting crosses were carried out as described previously (Gong and Golic,2003). The Abl stocks used were Abl¹,Abl⁴, Abl^Df(3L)st-j7 [from Dr Eric Libel(Liebl et al., 2003)], and Abl^EP3101 (Bloomington Stock Center). To minimize the potential effects of the genetic background, abi and Ablmutations (abi^P1, abi^P2, abi^KO,Abl⁴ and Abl^EP) were outcrossed with w¹¹¹⁸ for at least five generations. The UAS-GFP-Abi transgenic flies were generated by subcloning the GFP-Abi cDNA from pAc5-GFP-Abi(Huang et al., 2007) into pUAST. The following lines were obtained from the Bloomington Stock Center: UAS-Abl (8567), act5c-GAL4 (ubiquitous expression; 3954), elav^C155-GAL4 (pan-neuronal expression; 458) and ena^GC5 (8570). The 24B-GAL4 (muscle expression)line was provided by Dr Cheng-Ting Chien (Institute of Molecular Biology,Academia Sinica, Taipei, Taiwan). The scar^Δ37 and kette^C3-20 flies were obtained from Dr Eyal D. Schejer(Zallen et al., 2002) and Dr Christian Klämbt (Hummel et al.,2000), respectively.

Embryos were immunostained according to standard procedures(Patel et al., 1987). To study synapse morphology, third instar larvae were dissected for immunolabeling as previously described (Bellen and Budnik,2000). Bouton numbers on neuromuscular junctions (NMJs) were scored by counting distinct sphere-shaped synaptic terminals. Antibodies used were: rabbit anti-Abi (1:100) (Huang et al., 2007); mouse anti-BP102 (1:50), mouse anti-FasII (ID4, 1:50),mouse anti-Csp (1:40), mouse anti-Dlg (1:100) and mouse anti-Brp (nc82, 1:10)[Developmental Studies Hybridoma Bank, Iowa City, USA (DSHB)]; FITC-conjugated anti-Hrp (1:100, Jackson Laboratories); and rabbit anti-β-galactosidase(1:500, Chemicon). S2 and BG2-c2 cells were immunostained as previously described (Huang et al.,2007). Antibodies used were mouse anti-Ena (5G2, 1:50) (DSHB) and mouse anti-HA (1:1000, Babco). Rhodamine and Cy5-labeled donkey secondary antibodies were purchased from Jackson Laboratories. Confocal images were obtained using Leica TCS SP5 or TCS NT confocal microscopes.

Electrophysiological recordings were performed as previously described(Tsai et al., 2008).

Larvae were filleted and processed for transmission electron microscopy as described previously (Bellen and Budnik,2000). A total of six control (w¹¹¹⁸) and seven Abl^¼ boutons were measured. Synaptic vesicles in the clustered pool were defined as those <250 nm of the AZ T-bar. Measurements were quantified using ImageJ (NIH).

The culture of Drosophila S2 and BG2-c2 cells and transient transfection of plasmids into the cells were performed as previously described(Huang et al., 2007; Lin et al., 2007).

To date, no abi mutant strain has been reported. To determine the function of endogenous Abi, we generated a loss-of-function (LOF) mutant of abi by homologous recombination (see Materials and methods and Fig. S1A in the supplementary material). The isolated allele(abi^KO) is lethal when homozygous. We also identified two P-element insertion mutations (abi^P1 and abi^P2) in the abi gene locus. All four abi mutants analyzed (abi^P1, abi^P2,abi^KO and abi^Df) showed diminished Abi protein levels in homozygous embryos (Fig. 1A). Moreover, transheterozygotes with different combinations of the four alleles are lethal at different developmental stages. Specifically,the transheterozygotes abi^KO/Df,abi^P2/Df and abi^KO/P2 showed pupal lethality, suggesting that abi^KO and abi^P2 are strong LOF or null alleles. By contrast, over 60% of abi^P1/Df and abi^P1/KO animals survived to adulthood, suggesting that abi^P1 behaves genetically as a hypomorphic allele (Fig. 1B). The expression of an act5c-GAL4-driven GFP-Abi transgene could efficiently rescue the lethal phenotypes of abi mutants (see Fig. S1B in the supplementary material). Moreover, a line with a precise P-element excision from abi^P2 was found to have wild-type viability (data not shown). Our western blotting analysis confirmed the ectopic expression of GFP-Abi and the presence of no or only a low level of endogenous Abi protein in abi^KO/KO,abi^KO/P2 and abi^KO/Df adults (see Fig. S1C in the supplementary material). We concluded that abi^KO and abi^P2 are strong LOF alleles, and that abi^P1 is a hypomorphic allele.

Given that mammalian Abi family proteins are expressed at high levels in the developing nervous system (Proepper et al., 2007; Courtney et al.,2000; Grove et al.,2004), we expected Drosophila Abi to also be expressed in the nervous system. In our immunohistochemical study, we found that Abi is expressed primarily in the brain and ventral nerve cord (VNC) of the embryonic CNS (Fig. 1C,C′). Abi is specifically concentrated at the commissural and longitudinal connectives of the main axonal tracts. There was also some slight staining of Abi along the CNS midline (Fig. 1C′,arrowhead). At the larval stage, Abi is expressed in diverse regions of the CNS, including the neuropil area of the brain, axon fascicles in the VNC,photoreceptor cells in the eye disc and photoreceptor axons projecting through the optic stalk into the optic lobe, in contrast to the absence of immunoreactivity found in the abi^KO/P2 CNS(Fig. 1D,E). These results suggest that Abi plays a role in axonogenesis.

Therefore, we asked whether abi^-/- displays axonal patterning defects. Unexpectedly, we found no gross defects in BP102-positive axonal processes in the abi^KO/P2 embryonic CNS (see Fig. S2B in the supplementary material), raising the possibility that the maternal Abi protein might play a crucial role in CNS development. Thus, we generated abi germline clones(Chou and Perrimon, 1996), and found that embryos lacking both maternal and zygotic Abi(abi^MZ) displayed embryonic lethality and a catastrophic collapse of axonal patterning (see Fig. S2D in the supplementary material). Since the severe defects of the abi^MZ rendered the interpretation difficult, we used the maternal hypomorphic transheterozygous combination (abi^P1/Df) with the zygotic allele abi^P2 (referred to hereafter as maternal hypomorph, zygotic mutants) to assess the CNS defect. The embryos displayed deranged longitudinal connectives and commissural tracts (see Fig. S2C in the supplementary material). Together, these results clearly implicate maternal Abi in axonogenesis.

Previous genetic studies have found that Drosophila Abl is required for the inhibition of axon crossover at the CNS midline(Wills et al., 2002; Hsouna et al., 2003; Forsthoefel et al., 2005). This prompted us to investigate whether Abi is also involved in the regulation of midline crossing. We used anti-FasII antibody to label specific subsets of longitudinal axons. At embryonic stage 17, 9.6% and 31.25% of abi^KO/P2 and abi^KO/Df zygotic null embryos, respectively, exhibited midline crossing defects in ipsilateral axon fascicles(Fig. 1G,I). A higher penetrance (57.1%) of this phenotype was noted in the maternal hypomorph,zygotic mutants (Fig. 1H,I),suggesting that Abi might play a role in restricting specific longitudinal axons from crossing the midline.

Although there is a known biochemical interaction between Abl and Abi, no previous study has defined the genetic interaction between these two genes. The gene-dosage-sensitive genetic interaction analysis has been found to be effective in characterizing genes involved in Abl signal transduction networks (Lanier and Gertler,2000). We investigated whether Abl^-/-lethality could be affected by reducing the abi gene dosage. Surprisingly, the heterozygous abi genotype resulted in a dramatic increase in the survival of Abl^¼ progeny to adulthood (Fig. 2A). We found similar results in another Abl mutant(Abl⁴/Df(3L)st-j7) (data not shown), suggesting that abi acts as a genetic antagonist of Abl in the development of Drosophila.

Because both Abi and Abl contribute to CNS development, and the reduction of abi dosage suppresses Abl mutant lethality, we investigated the genetic interaction between abi and Abl in axonogenesis. Consistent with our survival analysis, abiheterozygosity markedly suppressed abnormal midline crossover in Abl^¼, from 37% of Abl^¼embryos having abnormal midline crossing to 18% of abi^P2/+;Abl^¼ embryos (Fig. 2B). This finding, which is similar to the findings previously reported for ena in the suppression of Abl axonogenesis phenotypes (Gertler et al.,1995), further suggests the possibility that Abi plays an opposing role to Abl in axonogenesis. Therefore, we investigated whether Ablmutations also modify abi mutant phenotypes in axonogenesis. Similarly, the presence of a single copy of the Abl gene suppressed the abi^KO/Df and abi^KO/P2midline-crossing phenotypes by approximately 53% and 69%, respectively(Fig. 2C). Together, these results suggest that abi and Abl seem to act against or oppose each other in the axonogenesis of embryonic CNS.

We noticed inactivity and a lack of coordination in abi and Abl flies. Performing negative geotaxis assays, we found significant reductions in locomotor activity in both abi and Abl mutants(see Fig. S3 in the supplementary material). Because the heterozygosity of abi suppressed axonal defects of Abl mutants, we tested whether abi mutations would also modulate the above-mentioned locomotion defects of Abl mutants. As expected, the presence of abi^-/+ increased the locomotor activity score of flies with Abl hypomorph mutations by more than 2-fold (see Fig. S3B in the supplementary material). However, caution should be used when interpreting these results, as locomotor abnormalities can be caused by a broad spectrum of developmental and/or functional defects, making it difficult to systematically examine the molecular mechanisms underlying this interesting phenotype of abi and Abl flies.

Because previous studies of mice have suggested that Abi and Abl proteins play a role in synaptic function (Moresco and Koleske, 2003; Grove et al., 2004; Proepper et al.,2007), we were prompted to study whether abi and Abl would genetically interact during synapse formation. We chose to use the Drosophila larval NMJ to study the genetic interactions during synapse formation (Collins and DiAntonio, 2007). We chose to study muscle 6/7 in abdominal segment A2 because no axonal targeting defects have been found in this region for either abi (our unpublished observations) or Abl mutants(Kraut et al., 2001). We first examined the NMJ morphology and electrophysiology function of abinull mutants and found no obvious phenotype for abi mutants. Both the number of synaptic boutons and synaptic transmission appeared to be unaffected by the mutations (see Fig. S4 in the supplementary material).

For the Abl mutants, however, the total synaptic bouton number(Fig. 3B) and satellite bouton formation (Fig. 5B, arrows)were significantly increased compared with w¹¹¹⁸ controls(Fig. 3A; Fig. 5A). There was a 42% and 67% increase in the normalized bouton number for Abl hypomorphic(Abl^4/EP) and null(Abl⁴/Df(3L)st-j7) larvae, respectively(Fig. 3E). To investigate whether a pre- or post-synaptic function of Abl is required for the synaptogenesis, we used the GAL4/UAS system(Brand and Perrimon, 1993) to ectopically express a neuronal or muscle Abl transgene in an Abl⁴/Df(3L)st-j7 mutant background. The presynaptic function of Abl appeared to be essential for synaptogenesis,because the selective expression of neuronal Abl (with an elav-GAL4driver), instead of muscle Abl (with a 24B-GAL4 driver),fully rescued the Abl⁴/Df(3L)st-j7 NMJ phenotypes(Fig. 3C-E). These results suggest that the presynaptic contribution of Abl is crucial for controlling the overgrowth of synapses in the NMJ. We also found that the Abl^-/- locomotion defect was alleviated by the re-expression of neuronal, but not muscle, Abl (data not shown). We concluded that the presynaptic function of Abl is crucial for synaptogenesis in Drosophila.

To test whether the synaptic overgrowth altered synaptic transmission in Abl^¼ mutants, we performed intracellular recordings from muscle 6 at segment A3 at the larval NMJs. The amplitude of both evoked excitatory junctional potentials (EJPs) and spontaneous miniature excitatory junctional potentials (mEJPs) remained unaffected in Abl^-/- larvae (Fig. 4A-D), but the frequency of mEJPs increased by 57% compared with the controls (Fig. 4B,E).

To investigate whether the altered synaptic transmission was accompanied by a synaptic structural abnormality, we stained NMJs with an antibody that recognizes Brp, which labels neurotransmitter-release sites at the active zone(AZ). We found no obvious morphological differences between Abl^¼ and controls(Fig. 5A,B). There was also no significant difference in the number of putative AZs per bouton area in Abl^¼ compared with controls (P=0.13)(Fig. 5G). In our examinations of other pre- and post-synaptic markers, including Csp, Futsch, FasII, GluRIIA and Dlg, we found no difference in immunoreactivity between the mutant and control NMJs (data not shown). These results suggest that the overall integrity of the synapse structure in the NMJ is not noticeably compromised in Abl mutants.

We conducted a transmission electron microscopy study to search for ultrastructural defects in Abl^-/-. However, general features of synapse structure, including presynaptic bouton morphology, the electron-dense AZs with T-bars and the structure of the subsynaptic reticulum(SSR), appeared to be unaffected in Abl^-/- synapses(Fig. 5D). Morphometric quantification of the number of synaptic vesicles clustered at the AZ, the number of AZs per section (Fig. 5C-G), the length of AZs and the overall size of synaptic vesicles(data not shown) also showed no significant differences between Abl^-/- and control synapses. Nevertheless, we found that the average density of the total synaptic vesicles was decreased by ∼50%in an Abl^-/- bouton compared with controls, although the vesicle numbers in the clustered pool were not notably affected(Fig. 5G). In addition, we frequently observed a number of enlarged but electron-clear vesicles near the T-bar in Abl mutants (Fig. 5F, arrow). As we did not observe an obvious increase in mEJP amplitude, it is likely that these large electron-clear vesicles contained fewer neurotransmitters than the electron-dense vesicles. Together, the decrease in overall synaptic vesicle numbers in the bouton areas might offset the increase in bouton numbers, in such a way that the EJP amplitude remains unchanged in Abl^-/- synapses.

Since the data above suggested that Abl and Abi might contribute to a common regulatory pathway in the developing nervous system, we investigated whether abi and Abl genetically interacted in NMJ development. Similar to the effects observed in axon guidance, removing one copy of abi moderately, but significantly, suppressed the Abl^-/- NMJ phenotypes(Fig. 6A), suggesting that Abi activity is required for synaptic overgrowth in Abl.

We and others have previously reported an association between Abi and Abl,and showed that this association promotes Abl-mediated phosphorylation of Ena in cell culture systems (Juang and Hoffmann, 1999; Tani et al.,2003). It is intriguing to note that ena also acts as a genetic antagonist of Abl in axonogenesis(Gertler et al., 1995) and its protein localizes to motor-axon terminals(Martin et al., 2005). To investigate whether ena acts cooperatively with abi to functionally antagonize Abl^-/- phenotypes, we examined whether double heterozygosity for abi and ena further suppressed Abl^-/- NMJ phenotypes. We found an additional suppression of NMJ overgrowth phenotypes in ena^-/+;Abl^-/-, abi^-/+ larvae(Fig. 6A). Together, these data lend support to the conclusion that Abi and Ena act in concert to antagonize the role of Abl in the development of the NMJ.

Previous studies have demonstrated that mutations of other SCAR/WAVE complex components (containing Scar, Abi, Kette, Sra-1 and HSPC300) also result in similar axonal abnormalities to those of abi mutants(Bogdan and Klambt, 2003; Qurashi et al., 2007; Schenck et al., 2003; Schenck et al., 2004). The similarity in the phenotypes prompted us to investigate whether the mutation of other components of SCAR/WAVE complex would also have a similar effect on inhibiting Abl synaptic phenotypes. However, the results are fairly divergent with regards to the inhibitory effect on Abl synaptic phenotypes. Similar to the findings in our study of abi^KO/+, scar heterozygosity(scar^Δ37/+) resulted in a significant decrease in NMJ bouton number in Abl larvae. However, a mutation in kette(kette^C3-20/+) appeared to be incapable of repressing the same phenotypes (Fig. 6B). These results might suggest that some, but not all, of the components of the pentameric SCAR/WAVE complex are involved in antagonizing the Abl-induced synaptic overgrowth.

Our results so far suggest that abi ena double heterozygosity can restrain the synaptic overgrowth of Abl mutants. Since both Abi and Ena proteins have been implicated in actin binding and reorganization(Krause et al., 2003; Innocenti et al., 2004), we asked whether F-actin structure was modulated by the overexpression of Abi or Ena in neuronal cells. The overexpresssion of either Abi or Ena alone in BG2-c2 neuronal cells redistributed peripheral F-actin to cytoplasmic aggregates wherever Abi or Ena had accumulated(Fig. 7A). However, Ena appeared to be more potent than Abi in redistributing F-actin to the cytoplasm(compare Fig. 7Ad-f with 7Aj-l). The cytoplasmic F-actin punta were efficiently dispersed and redistributed back to the cell periphery when Abl was co-expressed with either Abi or Ena(Fig. 7A). However, in cells co-expressing both Ena with Abl, Ena proteins were not completely relocalized to the cell periphery, but rather appeared to exist as cytoplasmic aggregates(Fig. 7Am-o), suggesting that Abl might affect the localization of Abi and Ena in neuronal cells differently. Further biochemical assays found Abi to be crucial for the association of Abl with Ena (data not shown). Immunocytochemical analysis also supported the co-existence of Abl-Abi-Ena complexes in cells(Fig. 7B; see Fig. S5 in the supplementary material). Together, these results suggest that Abl, through binding and modulating Abi and Ena activity, controls F-actin localization to cell periphery.

Because neurite outgrowth is regulated by the reorganization of peripheral actin (Schaefer et al., 2008),we investigated whether the Abl-Abi-Ena interaction affects neuritogenesis. We found that overexpression of Abi or Ena alone triggered the transformation of the extended biopolar neuronal cells into retracted cells (Fig. 7Ad-f and 7Aj-l). We also found that the retracted morphology of neuronal cells could be efficiently reversed by the co-expression of Abl with either Abi or Ena (Fig. 7Ag-i and 7Am-o). The striking morphological changes observed in neuronal cells prompted us to measure the changes in neurite outgrowth in cells overexpressing Abl, Abi or Ena, either singly or in pairs. In general, over 60% of the BG2-c2 cells displayed bipolar, neurite-like structures 4 days after plating onto a coverslip(Fig. 8A). We found a 3- to 60-fold decrease in the percentage of cells with bipolar morphology when overexpressing Abi or Ena alone (Fig. 8B). As expected, co-expressing Abl with Abi or Ena restored bipolar morphology (Fig. 8B),suggesting that Abl plays an opposing role to that of Abi and Ena in neuritogenesis. Since both Abi and Ena are physically associated with and functionally modulated by Abl, we speculated that Abi and Ena might act cooperatively to regulate the activity of Abl in neuritogenesis. If so, the co-expression of Abi and Ena might further suppress Abl function in neuritogenesis. We found that co-expressing both Abi and Ena with Abl could diminish neurite extension more than by overexpressing Abl with either Abi or Ena alone (Fig. 8B). To rule out the possibility that the changes in neurite extension were related to the levels of Abi, Ena and Abl protein expression, we conducted western blot analysis. We found comparable protein expressions in our experiments with different combinations of gene expression (see Fig. S6 in the supplementary material).

The in vivo role of Abi with respect to Abl has remained enigmatic. Abi was first identified as an Abl kinase substrate, functioning in modulating the transformation activity of oncogenic Abl in human cancers(Dai and Pendergast, 1995; Shi et al., 1995; Wang et al., 2007; Yu et al., 2008). Intriguingly, we and others have shown that Abi also functions as an activator of Abl kinase activity (Juang and Hoffmann, 1999; Tani et al.,2003; Lin et al.,2004; Leng et al.,2005; Maruoka et al.,2005). Moreover, the interaction of Abl and Abi can trigger an array of biochemical and functional changes in Abi, including protein phosphorylation, stability and subcellular localization(Huang et al., 2007), which might ultimately lead to the control of a particular biological process in vivo. Although both Abi and Abl proteins are highly expressed in the mammalian and Drosophila nervous systems, the role of Abi in modulating the function of Abl in developing nervous systems has remained unclear. In this investigation, we conducted genetic and functional studies to advance our understanding of how abi and Abl interact in vivo. To do this, we generated and characterized abi loss-of-function alleles for genetic and functional studies in Drosophila. Immunohistochemical analysis revealed that Abi is primarily expressed in the developing CNS. Consistent with this finding, our phenotypic analysis suggested that mutations in abi resulted in axonal guidance defects in the CNS. In an analysis of Abl mutants, we found Abl to be crucial for restricting synaptic overgrowth in the larval NMJ. Importantly, our further studies of the genetic interaction found a functional link between abi and Abl in axonogenesis and synaptogenesis. Moreover, Abi and Ena were found to cooperate in modulating the function of Abl in NMJ growth. Finally, based on additional cellular biology studies, we propose that the functional interactions between Abi, Ena and Abl might be mediated through the modulation of actin cytoskeleton reorganization.

Accumulated evidence suggests that the highly conserved actin-regulatory pathways are essential for synaptogenesis and synaptic plasticity(Coyle et al., 2004; O'Connor-Giles et al., 2008; Pawson et al., 2008; Rodal et al., 2008; Stewart et al., 2002; Stewart et al., 2005). Abi,Ena and Abl proteins are all involved in actin dynamics(Stradal and Scita, 2006; Lanier and Gertler, 2000). Using the Drosophila NMJ as a model system, we propose that the abi-ena-Abl interaction in synaptogenesis might be associated with actin cytoskeleton reorganization. In fact, several actin regulatory molecules associated with both Abl and Abi have been implicated in synaptic growth. For example, Wiskott-Aldrich Syndrome protein (Wasp) is a kinase substrate of Abl (Burton et al.,2005) and also a binding partner for Abi(Bogdan et al., 2005; Innocenti et al., 2005). The mutations in wasp result in phenotypes very similar to those present in Abl mutants, with synaptic overgrowth and hyperbranching at the NMJ (Coyle et al., 2004). Another example is that of Diaphanous (Dia), which also interacts with both Abl and Abi, and has recently been found to modulate synaptic growth of the Drosophila NMJ (Pawson et al.,2008). dia mutant heterozygotes have been found to be able to enhance the cellularization phenotype of an Abl maternal-null mutant (Grevengoed et al.,2003), suggesting that Dia might be involved in the regulation of Abl signaling for actin reorganization. Consistent with this idea, the interaction of Dia with Abi protein has been found to be important in regulating the formation and stability of cell-cell junctions in mammalian cells (Ryu et al., 2009). Future studies investigating whether and how Wasp and/or Dia can participate in Abl-Abi signaling for the regulation of Drosophila synaptogenesis could be interesting.

Besides Wasp and Dia, other actin regulators might also contribute to Abl-Abi signaling during nervous system development, for, as another study has suggested, Abl may be a key regulator in modulating different types and sites of actin polymerization within the cells(Grevengoed et al., 2003). Abi has been shown to play a key role in the activation of the SCAR/WAVE complex,which relays signaling from Rac1 to the Arp2/3 complex for actin cytoskeleton remodeling (Innocenti et al.,2004). Our genetic studies showed that the heterozygosity of scar, but not of kette, suppressed Abl NMJ phenotypes. A current model suggests that eliminating any component from the SCAR/WAVE complex induces the breakdown of other complex components and subsequently results in abnormal lamellipodia formation(Stradal and Scita, 2006). Our genetic study using the Drosophila NMJ as a model does not appear to fully support this idea. Our results suggested that only a subcomplex of SCAR/WAVE might be involved in synaptogenesis. In fact, recent studies have demonstrated that some components of the SCAR/WAVE complex might work outside the complex to regulate various biological processes, including neutrophil chemotaxis, cell motility and adhesion, and the formation of cell-cell junctions (Weiner et al.,2006; Pollitt and Insall,2008; Ibarra et al.,2006; Ryu et al.,2009). Thus, it is possible that Kette is not in a complex with Abi and Scar to modulate the function of Abl in NMJ growth. To explore this hypothesis, it will be important to examine the genetic interactions between abi and scar or kette in NMJ morphogenesis.

This study found strong in vivo evidence for an antagonistic relationship between Abl and Abi in axonogenesis and synaptogenesis. Supporting this model,one very recent study has demonstrated that Abl can inhibit the role of Abi in the engulfment of apoptotic cells in C. elegans(Hurwitz et al., 2009). Given that Abi is the Abl kinase substrate and that it also functions as an adaptor protein for Abl in regulating other downstream effectors(Juang and Hoffmann, 1999; Tani et al., 2003; Lin et al., 2004; Leng et al., 2005; Maruoka et al., 2005), it is feasible that Abi might act downstream of Abl in modulating NMJ growth. If so,the removal of both copies of abi could conceivably further suppress Abl^-/- NMJ phenotypes. Our preliminary morphological and functional data both suggest that the minor NMJ defects of Abl^-/- abi^+/- are further rescued in Abl^-/- abi^-/- mutants (data not shown). However, Abl^-/- abi^-/- double mutants showed early lethality and defects in axonal innervations, rendering the finding inconclusive. Further epistasis analysis combining abi and abl gain-of-function and loss-of-function mutations are needed to test this hypothesis.

Since our data suggest an antagonistic interaction between abi and Abl for the CNS and NMJ phenotypes(Fig. 2), we speculated that Abl heterozygosity would suppress the semilethal phenotype of abi mutants. Surprisingly, our preliminary data showed that the lethality of abi hypomorphic mutants (abi^P1/KOand abi^P1/Df) was further increased by Abl^+/- (data not shown). This result does not seem to support a general bidirectional antagonistic relationship between Abland abi for the biological processes involved during development. Thus, a complex genetic interaction network between Abl and abi might be present in development processes.

Another interesting issue is that the abi mutants did not display obvious defects in synaptic bouton number or synaptic transmission, although they exhibited midline crossing defects in the embryonic CNS. Because other members of SCAR complex, including Scar, Kette, Sra-1 and HSPC300, exhibit both CNS and NMJ phenotypes (Bogdan et al.,2004; Hummel et al.,2000; Qurashi et al.,2007; Schenck et al.,2003; Schenck et al.,2004), it is still possible that abi mutants might show minor morphological or functional abnormalities if different phenotypic characteristics are studied. Detailed morphological assays are required to investigate other phenotypic traits of the NMJ in larval or later developmental stages. Alternatively, one could reason that the roles of Abi in synaptic growth and axonal guidance are not exactly identical. Results similar to this finding have been observed for the loss of spastin, a gene enriched in axons and synaptic connections, as spastin mutants only exhibit NMJ but not CNS defects (Sherwood et al., 2004).

Our work also suggested that the synaptic overgrowth phenotypes in Abl mutants could be completely rescued by expressing Abl in the presynaptic nerve cells but not in the postsynaptic muscles, suggesting that the presynaptic Abl is more crucial than the postsynaptic population for Drosophila larval NMJ formation. However, studies in mammalian Abl and Arg (also known as Abl2) have shown that both proteins localize to the presynaptic terminals and dendritic spines of synapses in the hippocampal CA1 area (Moresco and Koleske,2003). Abl and Arg have also been shown to be essential for the agrin-induced clustering of acetylcholine receptors (AChRs) on the postsynaptic membrane of the mammalian NMJ, suggesting that Abl function is required in the postsynaptic region of the mammalian NMJ(Finn et al., 2003). However,these reports do not exclude the possibility that Drosophila Abl might also function in postsynaptic regions of the developing brain. The reason for this speculation is that the mammalian NMJ uses acetylcholine as the neurotransmitter, unlike the Drosophila NMJ, which uses glutamate as a transmitter. Since acetylcholine receptors also function in the developing brain of Drosophila(Yasuyama et al., 1995), it would be important to investigate whether Drosophila Abl also plays a role in the postsynaptic region of the neurons, where acetylcholine receptors are expressed.

In conclusion, our genetic studies in Drosophila suggest that Abi and Abl play opposing roles in axonogenesis and synaptogenesis. This conclusion is further supported by a series of biochemical, immunocytochemical and morphological studies in cultured cells. These findings offer new insights into the functional interaction between Abl, Abi and Ena in nervous system development.