The ability of neuronal growth cones to be guided by extracellular cues requires intimate communication between signal
transduction systems and the dynamic actin-based cytoskeleton at the leading edge. Profilin (chickadee), a small, actin-binding protein,
has been proposed to be a regulator of the cell motility machinery at leading edge membranes. However, any requirement it may have in
the developing nervous system has been unknown. Profilin associates with members of the Enabled family of proteins,
suggesting that Profilin might link Abl function to the cytoskeleton.
In a genetic screen in Drosophila to identify genes required for the correct navigation and
outgrowth of motoneuron growth cones two alleles of a
stranded (sand) mutation were recovered in which motor growth cones arrest before reaching their final targets. The molecular genetic analysis reveals that stranded alleles are zygotic lethal mutations in Profilin. In vitro experiments
confirm that axon extension is impaired in Profilin mutants. Moreover, phenotypic comparisons and genetic interactions
between chic and abl mutants support the notion that Profilin and Abl cooperate to promote axon
extension. Genetic analysis in Drosophila has been used to
demonstrate that mutations in Profilin (chickadee) and Abl (abl) display an identical growth cone arrest phenotype for axons
of intersegmental nerve b (ISNb). Moreover, the phenotype of a double mutant suggests that these components function
together to control axonal outgrowth (Wills, 1999a).
Genetic analysis of growth cone guidance choice points in Drosophila has identified neuronal receptor protein tyrosine
phosphatases (RPTPs) as key determinants of axon pathfinding behavior. The Drosophila Abl
tyrosine kinase functions in the intersegmental nerve b (ISNb) motor choice point pathway as an antagonist of the RPTP
Dlar. The function of Abl in this pathway is dependent on an intact catalytic domain. The Abl
phosphoprotein substrate Enabled (Ena) is required for choice point navigation. Both Abl and Ena proteins associate with
the Dlar cytoplasmic domain and serve as substrates for Dlar in vitro, suggesting that they play a direct role in the Dlar
pathway. These data suggest that Dlar, Abl, and Ena define a phosphorylation state-dependent switch that controls growth
cone behavior by transmitting signals at the cell surface to the actin cytoskeleton (Wills, 1999b).
The reciprocal catalytic activities of a tyrosine kinase and phosphatase predict that a reduction in kinase activity within the Dlar pathway might suppress the Dlar motor axon phenotype. In Dlar mutant embryos, subsets of axons derived from the intersegmental nerve route (ISN), called ISNb and ISNd, fail to enter adjacent muscle target domains just outside the ventral nerve cord. Instead, Dlar mutant ISNb and ISNd axons follow the ISN toward dorsal targets (the bypass phenotype. Since abl loss of function is known to disrupt the outgrowth of ISNb, the Abl tyrosine kinase is an excellent candidate for a role in Dlar signaling. Therefore, various genetic backgrounds were examined in which homozygous Dlar mutations were combined with mutations in a single allele of abl. Reduction of abl of up to half the normal gene dose has a profound effect on the penetrance of the Dlar motor axon guidance phenotype, suppressing the Dlar phenotype up to 10-fold; for example, ISNb bypass in Dlar mutants is reduced from 38% to 4% in abl heterozygote mutants (Wills, 1999b).
Western blot analysis shows that endogenous Abl protein binds specifically to the full-length Dlar cytoplasmic domain (GST-Dlar D1-D2). The association of Dlar and Abl in cell extracts is consistent with a direct functional relationship between the two proteins. However, the binding could depend on other factors present in the crude extract. Therefore, the association of purified recombinant Abl protein with Dlar fusion proteins was examined in the absence of other Drosophila proteins. Recombinant Abl binds to Dlar with somewhat less specificity than does the Abl endogenous to S2 cells. Purified mammalian v-Abl binds to Dlar under the same conditions, with a profile of specificity very similar to that of Drosophila Abl. Since v-Abl represents only the kinase and SH2 domains of Abl, these domains appear sufficient to mediate Dlar binding. As further evidence of direct physical interactions between Abl and the Dlar D2 domain, kinase assays reveal that Drosophila Abl phosphorylates GST-Dlar D2 in vitro. In addition to the Dlar D2 domain, Drosophila Abl can weakly phosphorylate the D2 domain of another receptor tyrosine kinase, Protein tyrosine phosphatase 69D (Ptp69D); this is interesting, since Ptp69D is tyrosine phosphorylated in S2 cells. The physical interactions between Abl and Dlar support a model whereby both proteins function in the same signaling pathway. Furthermore, the phosphorylation of the D2 domain in vitro raises the intriguing possibility that d-Abl activity regulates Dlar function in vivo (Wills, 1999b).
The contrast between the abl and Dlar phenotypes and the suppression of the Dlar phenotype by abl alleles suggest that Abl and Dlar play functionally antagonistic roles in ISNb development. This hypothesis makes a simple prediction: gain of function in Abl should result in a phenotype similar to loss of Dlar. Therefore, the GAL4 expression system was used to target high-level expression of wild-type Abl to postmitotic neurons and then the development of motor axon pathways was examined. With three independent neural specific GAL4 drivers, in combination with an abl cDNA under the control of the GAL4 upstream activator sequence (UAS), GAL4-dependent phenotypes were observed. When wild-type Abl is overexpressed, ISNb axons bypass their ventral target muscles in a manner indistinguishable from that of the ISNb phenotype observed in Dlar mutants. The kinase activity of Abl has been shown to be necessary for its role in ISNb neurons (Wills, 1999b).
Since Ena acts as a genetic antagonist of Abl, it was reasoned that loss of Ena should resemble gain of Abl. ISNb bypass phenotypes are seen in all ena mutant combinations. Two types of ISNb phenotypes are observed in ena mutants: (1) failure of ISNb to enter the ventral muscles after a successful defasciculation (characteristic of embryos lacking Dlar alone), and (2) failure of ISNb axons to defasciculate from the ISN pathway (characteristic of embryos lacking multiple phosphatases. In addition, the frequency of ISNb bypass in strong ena mutants is twice that observed in the strongest Dlar alleles. These observations may indicate that Ena acts as a point of convergence for multiple inputs in the ISNb guidance mechanism. Ena family members share a conserved domain structure, including an N-terminal EVH1 domain that mediates binding to Zyxin and Listeria ActA, a proline-rich region that supports associations with Profilin and SH3 domains, and a C-terminal EVH2 domain that promotes multimerization. Mutations are available that specifically disrupt either the EVH1 or the EVH2 domains of Ena. Mutations in either domain display highly penetrant ISNb bypass, demonstrating a requirement for both domains in the guidance mechanism. Although Ena is restricted to axons in the developing nervous system late in embryogenesis, it is expressed broadly prior to germ band retraction. To confirm that neuronal Ena function is necessary for ISNb choice point navigation, wild-type ena cDNA was expressed under neuronal GAL4 control in an ena mutant background. Neural specific ena expression attenuates the ISNb phenotype significantly. If the quantity of Ena protein is rate limiting in wild-type ISNb axons, one might expect Ena overexpression to disrupt ISNb guidance. However, no ISNb phenotypes are observed, even when UAS-ena is combined with the strongest neural driver P[elav-GAL4] (Wills, 1999b).
The genetic relationship between Abl and Dlar and the requirement of Ena function for ISNb target entry suggest that Ena might act in the Dlar signaling pathway. To test this model, it was asked whether Ena associates with the cytoplasmic domain of Dlar. Endogenous Ena protein associates with a Dlar full-length cytoplasmic domain (GST-Dlar D1-D2) or with D2 alone but not comparably with wild-type D1. Since Abl is known to associate with Ena, and since binding between Abl and Dlar has been demonstrated, it is possible that Ena binding to Dlar requires Abl or additional proteins. Purified Ena has been shown to bind to the Dlar cytoplasmic domain. In both extract and recombinant protein binding assays, Ena shows only weak association with DPTP10D. However, Ena binds effectively to the D2 domain of Ptp69D. The preferential binding of Ena to the D2 domains of Dlar and Ptp69D, as compared with the D1 domains of the same RPTPs, suggests that these interactions are specific. The parallel between Dlar and Ptp69D binding is interesting, given the published observation that Ptp69D is required for ISNb guidance and can partially substitute for Dlar in vivo. Furthermore, the nature and penetrance of ISNb defects in ena mutants suggest that Ena may function downstream of multiple inputs (Wills, 1999b).
The relationships between Abl, Ena, and Dlar in motor axon guidance suggest a model whereby Abl and Dlar compete for shared substrates to regulate growth cone behavior. Although the Dlar cytoplasmic domain was previously shown to encode an active PTP domain, using artificial phospho-peptide substrates in vitro, no physiological substrates have been identified. Since nearly all of the tyrosine phosphatase activity of LAR family RPTPs resides in the D1 domain, the ability of the GST-Dlar D1 fusion protein to dephosphorylate purified Drosophila Abl or Ena proteins after these proteins have been phosphorylated with recombinant d-Abl was examined. Incorporated 32P is rapidly released from both Abl and Ena after addition of wild-type GST-Dlar D1 but not after addition of the catalytically inactive C-to-S mutant GST-Dlar D1 fusion protein. These results suggest that the bacterially expressed GST-Dlar protein is correctly folded and that Drosophila Abl and Ena are both potential Dlar substrates. However, because PTPs are known to be promiscuous in vitro, additional experiments will be necessary to determine whether Abl and/or Ena are targets for Dlar activity in vivo (Wills, 1999b).
Drosophila Roundabout (Robo) is the founding member of a conserved family of repulsive axon guidance receptors that respond to secreted Slit proteins. Little is known about the signaling mechanisms that function downstream of Robo to mediate repulsion. Genetic and biochemical evidence is presented that the Abelson (Abl) tyrosine kinase and its substrate Enabled (Ena) play direct and opposing roles in Robo signal transduction. Genetic interactions support a model in which Abl functions to antagonize Robo signaling, while Ena is required in part for Robo's repulsive output. Both Abl and Ena can directly bind to Robo's cytoplasmic domain. A mutant form of Robo that interferes with Ena binding is partially impaired in Robo function, while a mutation in a conserved cytoplasmic tyrosine that can be phosphorylated by Abl generates a hyperactive Robo receptor (Bashaw, 2000).
Abl and Ena are complementary components of the signaling machinery downstream of the Robo repulsive axon guidance receptor. Genetic interactions indicate that loss of ena function partially disrupts Slit- and Robo-mediated repulsion from the midline. Limiting or removing ena function enhances partial loss-of-function robo phenotypes and suppresses robo gain-of-function phenotypes. In contrast, reduction of abl has the opposite consequence, suppressing the effects of a partial loss of robo function, while panneural overexpression of Abl antagonizes Robo function, leading to a phenotype resembling that of robo mutants (Bashaw, 2000).
Both Abl and Ena bind directly to Robo's cytoplasmic domain in vitro and Robo can act as a substrate for Abl kinase activity in vitro and in cell culture. Robo and Ena also show in vivo physical interactions. Furthermore, cytoplasmic domain mutants that reduce Ena binding to Robo result in impaired ability to rescue robo loss of function, while a Y-F mutation in a conserved tyrosine that can be phosphorylated by Abl in vitro has the opposite consequence, generating a hyperactive Robo receptor. These genetic and biochemical data support a model in which Abl and Ena play direct and opposing roles in the transmission of Robo's repulsive signal (Bashaw, 2000).
The implication of Ena in repulsive axon guidance is somewhat surprising in light of the previous results from the pathogen Listeria monocytogenes indicating that Mena is required for Listeria's actin-polymerization dependent motility. The Listeria data, together with the in vitro effects on actin of the Ena/VASP proteins has frequently been interpreted to suggest that Ena/VASP proteins function to promote actin polymerization, thereby promoting motility. On the contrary, the results presented here indicate that Ena is partially required for axon repulsion from the midline. These data suggest that Ena may have the opposite function, namely, to inhibit forward growth cone motility at sites where Robo encounters Slit (Bashaw, 2000).
In a companion paper (Bear, 2000), an independent study in mammalian cell culture has reached a similar conclusion. By expressing a multimerized EVH1 domain binding site attached to specific subcellular localization sequences, Ena/VASP family members can be efficiently targeted to different areas of cultured fibroblasts. This system has allowed a direct examination of the role of Ena/VASP proteins in cell motility. Surprisingly, when Ena/VASP proteins are directed away from the cell membrane, using a mitochondrial targeting sequence, the cells actually migrate more quickly. Conversely, targeting Ena/VASP proteins to the membrane, or overexpressing Mena, leads to a dose-dependent decrease in the rate of cell migration. A major conclusion of this study is that Ena/VASP proteins function in part to decrease the rate of whole cell motility. Whether Ena/VASP proteins achieve the observed in vivo effects on whole cell and growth cone motility by stimulating or inhibiting actin polymerization awaits future investigation (Bashaw, 2000).
While the dosage-sensitive genetic interactions between ena and robo support a role for Ena in midline repulsion, Ena clearly can not explain all of Robo's repulsive output. Indeed, although mild midline crossing defects are observed in ena mutants, on the whole, Robo-mediated repulsion works fairly well in the absence of Ena. In this light, it is perhaps not surprising that the Robo DeltaCC2 mutant receptor (in which the Ena binding site is deleted) still provides some repulsive activity and can partially rescue robo loss-of-function mutants. These results indicate that there must be other proteins that function downstream of Robo to mediate repulsion. One would predict that simultaneously removing ena and the as yet unknown additional factors would reveal stronger disruptions of midline repulsion (Bashaw, 2000).
Thus, Ena is only part of what must be a more complex repulsive output from Robo. Ena helps strengthen the output (perhaps by locally putting the break on the actin-based motility machinery), but is only part of the output. In this light, it is interesting to note that Robo2 also binds Slit and mediates repulsion (albeit apparently more weakly than Robo), but Robo2 does not have the Ena binding site and does not bind Ena (J. Simpson, personal communication to Bashaw, 2000).
An important question for future studies concerns whether Ena is always docked on Robo, or alternatively, whether Slit binding to Robo leads to the recruitment of Ena to Robo's cytoplasmic domain. From what is known about other receptor systems, this second alternative seems more likely, but it remains an open question and needs to be directly tested (Bashaw, 2000).
Genetic analysis shows that Abl antagonizes Robo-mediated repulsion. The two most likely possibilities are that Abl functions to antagonize this pathway by phosphorylating Robo or by phosphorylating Ena. Three results argue in favor of a direct interaction with Robo. (1) Certain kinds of dose-dependent genetic interactions between abl and robo are observed that are not observed between abl and ena, suggesting that the Abl and Robo proteins might directly interact. (2) Biochemical experiments have shown that Abl can directly phosphorylate Robo's cytoplasmic domain at one or more tyrosine residues. (3) A Y-F mutation in a conserved tyrosine that can be phosphorylated by Abl in vitro generates a hyperactive Robo receptor. Taken together, these genetic and biochemical data suggest that it is the dephosphorylated form of Robo that is most active (Bashaw, 2000).
How might Abl normally regulate the output of Robo signaling? Abl-mediated phosphorylation might normally modulate the output of Robo signaling. Alternatively, this phosphorylation might participate more directly in the ligand-gated signal. It is interesting to speculate that it is the binding of Robo to its ligand Slit that triggers dephosphorylation, and that this in turn activates the repulsive response (Bashaw, 2000).
The CNS-specific receptor protein tyrosine phosphatases (RPTPs) RPTP10D and 69D are candidates to be additional factors that contribute to Robo repulsion. Simultaneous removal of these two RPTPs results in substantial ectopic midline crossing, and the double mutant shows dose-sensitive genetic interactions with slit. Whether these two phosphatases interact directly with Robo and whether their phosphatase activity is required for their observed roles in repulsion await future investigation (Bashaw, 2000 and references therein).
In the model presented above, it is attractive to speculate that these two RPTPs function in opposition to the Abl kinase activity by directly dephosphorylating Robo upon Robo's interaction with Slit. Interestingly, the other two Robo family members in Drosophila (Robo2 and Robo3; J. Simpson, personal communication to Bashaw, 2000) share the phosphorylation sites in Robo that are phosphorylated by Abl in vitro. In addition, genetic interactions are observed between the RPTPs and Robo2. Together these observations suggest that perhaps a common mechanism is employed to regulate the signaling output of the three Robo receptors. It will be of interest to determine the in vivo significance of the conserved tyrosine phosphorylation sites in the three Robo receptors. The future elucidation of the events set in motion by ligand binding will require the development of cell culture systems that will allow analysis of the phosphorylation state and cytoplasmic domain associations of the Robo receptors before and after Slit stimulation (Bashaw, 2000).
In addition to their function during Robo signaling shown here, it is clear that both Abl and Ena function in multiple guidance signaling pathways, and thus that they are not committed to repulsion downstream of Robo. In the nematode C. elegans, ena acts as a suppressor of the axon migration defects associated with ectopic expression of the UNC5 repulsive Netrin receptor. This raises the possibility that ena functions downstream of diverse repulsive guidance receptors. In Drosophila, during motor axon pathfinding, ena and abl play roles in ISNb choice point control. Overexpression of abl or loss of ena generates an ISNb 'bypass' phenotype, where the ISNb fails to defasciculate and branch off at the appropriate location to enter its muscle target region. This phenotype is also observed in mutations in Dlar, the gene encoding a receptor protein tyrosine phosphatase (RPTP). Mutations in all three of these genes (ena, abl, and Dlar) give rise to only partially penetrant ISNb guidance phenotypes and appear to modulate guidance decisions at this choice point. At the midline, mutations in the genes encoding the ligand (Slit) and a key receptor (Robo) have strong and highly penetrant midline guidance phenotypes. In contrast, mutations in the genes encoding Ena, Abl, and RPTP10D and RPTP69D on their own have weaker and less penetrant phenotypes. This is consistent with Abl and the RPTPs modulating Robo receptor output, and Ena mediating only part of Robo output (Bashaw, 2000 and references therein).
If this same logic is applied to the motor axon ISNb choice point, then it is likely that some of the key components are still missing. At present, the only gene with a nearly 100% penetrant bypass phenotype at this choice point is sidestep. Side is an Ig superfamily transmembrane protein that is expressed on muscle surfaces and appears to function as an attractive ligand for motor axons. The Side receptor is not known. Whether the key receptor is the Side receptor or not, it is likely that the major growth cone receptor for the ISNb choice point has not yet been identified (Bashaw, 2000 and references therein).
In this context, it is tempting to speculate, by analogy with the proposed model for Robo signaling, that at the ISNb motor axon choice point, DLAR and Abl play complementary roles in modulating the output activity of the hypothetical guidance receptor, while Ena functions as part of the receptor output. In this way, the two guidance decisions -- to cross or not to cross the midline, and to fasciculate or defasciculate from other motor axons -- use different signals on the outside of the growth cone, but similar signaling and regulatory mechanisms on the inside. It is suggested that once the signal crosses the membrane, in both cases the output is regulated in opposing directions by Abl vs. one or more RPTPs, and that the output is partially mediated by Ena. This model provides a unifying way of viewing signal transduction during these two different guidance decisions. It will be interesting in the future to see to what degree this model holds up in terms of both the role of phosphorylation in modulating receptor output, and the role of Ena in mediating part of repulsive signaling (Bashaw, 2000).
Drosophila capulet (capt), a homolog of the adenylyl cyclase-associated protein that binds and regulates actin in yeast, associates with Abl in Drosophila cells, suggesting a functional relationship in vivo. A robust and specific genetic interaction is found between between capt and Abl at the midline choice point where the growth cone repellent Slit functions to restrict axon crossing. Genetic interactions between capt and slit support a model where Capt and Abl collaborate as part of the repellent response. Further support for this model is provided by genetic interactions that both capt and Abl display with multiple members of the Roundabout receptor family. These studies identify Capulet as part of an emerging pathway linking guidance signals to regulation of cytoskeletal dynamics and suggest that the Abl pathway mediates signals downstream of multiple Roundabout receptors (Wills, 2002).
Previous studies in Drosophila have shown that although Abl, Ena, and Profilin proteins are expressed broadly during embryogenesis, at later embryonic stages they accumulate at highest levels within the developing nervous system. Capt is abundantly expressed in early stage embryos (e.g., stage 4), consistent with its documented role in oogenesis. In addition to expression in mesoderm and developing gut epithelia, Capt is abundant in the ventral nerve cord (VNC) at stages 12 and 13 when axon pathways are pioneered within the CNS. At late stages (stages 16 and 17), when the last axon pathways are maturing and synaptogenesis is beginning, Capt preferentially accumulates within the VNC. Therefore, Capt, Abl, Ena, and Profilin are coexpressed in neurons at stages important for axonal development (Wills, 2002).
Capt and Abl are able to associate in a physiological setting. Using monoclonal antibodies (mAbs) specific to Abl, it is possible to immunoprecipitate (IP) endogenous Abl from Drosophila S2 cell lysates. When IPs of Abl were analyzed by SDS-PAGE and subsequent Western blot with anti-Capt antisera, endogenous Capt was detected as a protein that coprecipitates with Abl. For further confirmation of Capt protein association, S2 cells were transfected with a cDNA construct encoding full-length capt with a hemoagglutinin (HA)-epitope tag. Anti-HA antibody Western blots of Abl IPs detected a protein of the molecular weight expected for the tagged version of Capt (Wills, 2002).
Genetic experiments suggest that Abl interacts with a number of actin regulatory proteins to control cytoskeletal assembly. Given the functional redundancy observed between CAP and Profilin in yeast, it was thought that Capt and Profilin might participate in some form of protein complex regulated by the Abl kinase. S2 cells were transfected with full-length Drosophila Abl (dAbl), Drosophila Src64 (dSrc), or the truncated mammalian v-Abl and then Capt immunoprecipitations were assayed with anti-Profilin (Chic) and anti-actin antibodies. No significant binding of Capt and Profilin were seen in cells transfected with dSrc or v-Abl or in untransfected controls where endogenous dAbl is expressed at very low levels. However, an association of Capt with Profilin and with actin was observed when dAbl was elevated, suggesting a model where Abl, Capt, and Profilin function together in a cytoskeletal protein complex (Wills, 2002).
The expression and interactions of Capt protein raised the question of whether Capt contributes to the function of the Abl pathway during nervous system development. However, examination of many independent capt allelic combinations that remove zygotic expression without affecting other genes nearby revealed no defects in the embryonic CNS. This is attributed to the large maternal supply of Capt protein visible in the early embryo. Unfortunately, like Profilin null mutations (chickadee), capulet null alleles completely block oogenesis, preventing the use of germline mosaics for the study of zygotic phenotypes in the absence of maternal expression. However, because strong zygotic phenotypes can be induced when mutations in various Abl pathway components are combined with mutations in Abl (e.g., disabled), it was reasoned that zygotic functions of capt might be revealed through genetic interactions (Wills, 2002).
Among the strongest genetic interactions are synthetic phenotypes that arise in transheterozygotes, which lack only one copy of each interacting locus. Heterozygotes that lack one copy of capt or Abl alone show no detectable CNS phenotypes when compared to wild-type strains. However, combination of one capt and one Abl allele results in a distinct axon pathfinding defect (Wills, 2002).
Axons in the Drosophila embryonic CNS are organized into two major groups: longitudinal pathways that extend along the anterior-posterior axis and commissural pathways that carry contralateral projections across the midline. The midline, composed of specialized glial cells, acts as an organizing center that provides secreted growth cone attractants (Netrins) to build commissural pathways and a secreted repellent (Slit) to prevent inappropriate midline crossing. Subsets of longitudinal axons that depend on Slit to maintain their ipsilateral trajectories can be visualized specifically at late embryonic stages (stage 17) with the anti-Fasciclin II (FasII) antibody mAb 1D4; these FasII-positive axons never cross the midline in wild-type embryos (Wills, 2002).
In contrast to wild-type, capt-Abl transheterozygotes display consistent axon guidance defects at the CNS midline. In these double mutants, ipsilateral axon fascicles now ectopically cross, primarily from the most dorsal-medial MP1 pathway. An allelic series of this capt-Abl synthetic phenotype is seen across many different transallelic combinations, showing that the effect is independent of genetic background. No gross defects in the number or fates of postmitotic neurons were detected in any capt-Abl mutants. Although temporal delays were sometimes observed, capt-Abl transheterozygotes did not show any lasting defects in embryonic motor axon pathways (Wills, 2002).
To test whether the midline axon guidance function for capulet is dependent on Capt expression in postmitotic neurons, a wild-type capt transgene was expressed under control of P[elav-GAL4] in a strong capt-Abl background; a 15-fold rescue of the capt-Abl phenotype was observed. Interestingly, a parallel rescue experiment using an N-terminal deletion removing the putative Capt adenylyl cyclase-associated domain provides only a 2.7-fold rescue under the same conditions, despite the fact that the same transgene fully rescues captacuE636 to viability. Thus, capulet and Abl cooperate specifically during midline axon guidance (Wills, 2002).
The failure of the midline gatekeeper function in capt-Abl transheterozygote embryos suggested that Capt might function in the repellent pathway downstream of Slit. To test this genetically, transheterozygotes lacking one allele of capt and one allele of slit were examined. These mutants show a significant increase in the number of midline crossing errors compared to controls. This genetic interaction is seen consistently with multiple alleles of capt. Thus, capt and slit cooperate during midline guidance (Wills, 2002).
To further test the model that capt acts in the repellent pathway, the system of receptors was examined. However, examination of single gene mutations might not be sufficient. This is because the response to Slit is mediated by multiple receptors: Robo, Robo2, and Robo3. Indeed, capt transheterozygotes lacking single alleles in robo, robo2, or robo3 alone show little if any midline phenotype. Yet, when capt alleles are combined with double mutations lacking one copy of robo and robo2 simultaneously, a phenotype almost 2-fold greater than that seen in the robo,robo2 heterozygous embryos is observed. Interestingly, capt/+ does not enhance the phenotype of robo,robo3 heterozygotes, which is already quite strong (Wills, 2002).
As capt activity is further reduced, the interaction with robo2 gets stronger; mutants lacking two copies of capt and one copy each of robo2 and robo3 display penetrant midline phenotypes. Since these allelic combinations are the most severe, they were used for more detailed phenotypic analysis. For example, since the repulsion of growth cones at the midline is dependent on the presence of the midline glia, which secrete the Slit repellent, the midline glia in these mutants were examined with anti-Wrapper antibody, which specifically stains the surface of these glial cells. Midline glia are present in capt-robo2,robo3 mutants, even where axons inappropriately crossed the midline. The first axons in the MP1 fascicle were examined just as they pioneer the ipsilateral pathway early in CNS development. At stage 12, the posterior corner cell (pCC) extends its axon along an anterior trajectory parallel to the midline in order to pioneer the most medial Fasciclin II-positive (MP1) pathway. In capt-robo2,robo3 mutants, pCC axons were sometimes found that had turned toward and crossed the midline at this early stage. This phenotype is similar to that seen in robo alleles (Wills, 2002).
Previous studies of Abl function during midline guidance led to a model where Abl acts to antagonize Robo signaling. However, analysis of capt-Abl transheterozygotes suggests that Abl might play a dual role and also be required for restriction of midline crossing. Consistent with this prediction, examination of several Abl homozygotes reveals an allelic series of midline crossing phenotypes identical to those seen in capt-Abl transheterozygotes. Expression of a wild-type Abl transgene under its endogenous promotor in a strong mutant background rescues the midline crossing phenotype, as does expression of Abl specifically in neurons; however, a kinase-dead transgene was unable to rescue the defect. Like other aspects of Abl function, the midline crossing defects in Abl mutants can be suppressed by dose reduction of its substrate protein Ena or by loss of the receptor protein tyrosine phosphatase Dlar. These observations demonstrate that Abl is required for inhibiting the passage of ipsilateral axons across the midline and suggest that the role of Abl is more complex than previously appreciated (Wills, 2002).
Since analysis of Abl loss-of-function would predict cooperation between Abl and other genes in the repellent pathway, genetic interactions in embryos transheterozygous for Abl and either slit or combinations of mutations in different roundabout genes (ie. slit/+;Abl/+ or robo,robo2/+,+;Abl/+) were assayed. Surprisingly, these embryos displayed striking midline phenotypes far stronger than control genotypes. For example, slit2/+;Abl2/+ transheterozygotes show a 24-fold enhancement of the slit2/+ phenotype. This experiment strongly supports the model that Abl acts positively in the Slit pathway, consistent with the phenotypes of Abl homozygotes and of all the capulet genetic interactions observed (Wills, 2002).
The network of genetic interactions observed suggests that the Abl pathway is involved in signaling downstream of multiple Robo-family receptors. However, previous studies have shown Abl binding to the Robo cytoplasmic domain in vitro is dependent on a peptide motif (CC3) that is not present in Robo2 or Robo3. It was necessary to have an in vivo test for Abl-Robo interactions to explore this issue. Since Abl appears to act in both positive and negative capacities at the embryonic midline, an alternative genetic assay was used to evaluate Abl interaction with the robo gene family. When wild-type Abl is overexpressed in the developing compound eye, under the control of a synthetic glass promotor (GMR-GAL4), a mild rough-eye phenotype was observed. Thus, this Abl phenotype was tested for interactions with various UAS-Robo transgenes (Wills, 2002).
As predicted from loss-of-function analysis, while expression of wild-type Robo alone has little, if any, effect on retinal patterning, the combination of Abl and Robo causes a striking increase in the severity of the Abl gain-of-function eye phenotype. Thus, Robo serves as an enhancer of Abl activity in this kinase-dependent assay. This is also true of Robo2 and of Robo3. These data support the hypothesis that all Robo receptors can engage the Abl signaling pathway. So, is this in vivo interaction dependent on the Robo domains previously shown to recruit Abl and Ena proteins? Interestingly, neither deletion of CC2 nor deletion of CC3 was found to attenuate the Abl-Robo interaction. A UAS-robo transgene lacking the motif CC1 did show a reduction in eye phenotype when combined with UAS-Abl, but the difference was slight (Wills, 2002).
To confirm that Abl can interact with Robo in a CC3 domain-independent fashion during axon guidance, embryos that overexpress Abl and either wild-type Robo(+) or mutant Robo(DeltaCC3) were examined in postmitotic neurons. Abl gain-of-function alone generates two axon guidance phenotypes: (1) ISNb motor axon bypass of ventral target muscles and (2) ectopic midline crossing. Interestingly, coexpression of Abl and either Robo(+) or Robo(DeltaCC3) dramatically enhances the ISNb axon phenotype; however, there was no effect on midline crossing in any of these genotypes. Thus, in vivo, Abl is capable of a functional interaction with all three Robo receptors via some novel mechanism. However, the midline guidance system is specifically refractory to a simultaneous increase in Abl and Robo activities, perhaps due to the dual role of Abl in this context (Wills, 2002).
This study provides compelling evidence that a member of the adenylyl cyclase-associated protein (CAP) family plays a role in the accurate navigation of developing axons. Phenotypic analysis of double mutant embryos demonstrates that Capt cooperates with Abl, Slit, and multiple Roundabout receptors to prevent the inappropriate traffic of axons across the midline choice point. Consistent with published data on the relative contribution of Robo2 and Robo3 to midline repulsion, it has been found that capt and Abl show stronger interactions with robo,robo2 double mutants; however, Abl does appear to interact with all three receptors. The genetic and biochemical interactions observed suggest both that Capt functions directly in the Abl pathway and that this cytoskeletal regulatory pathway is involved in the repellent response to Slit (Wills, 2002).
Detailed studies of the prototypical growth cone repellent CollapsinI/Semaphorin3A have shown that the repellent response involves a collapse of the leading edge structures supported by actin cytoskeleton. Similar results have been seen for members of the Ephrin and Slit protein families. The fact that repellents promote a net disassembly of actin polymer arrays favors the simple model that repellent signaling antagonizes the actin assembly process (Wills, 2002).
Studies of CAP homologs from yeast, Dictyostelium, mouse, pig, and human suggest that the C-terminal actin binding domain acts to sequester monomers to prevent actin polymerization. More recent studies also suggest that human CAP promotes actin disassembly and monomer recycling through interactions with the actin-depolymerizing factor Cofilin. Consistent with an inhibitory role for CAP-family members, studies of epithelial development and oogenesis in Drosophila demonstrate that Capt functions to suppress the hyperassembly of actin microfilaments. Interestingly, a similar function has been ascribed to Abl and Arg during neurogenesis in the mouse. Thus, a model is favored where Abl helps to recruit and regulate CAP activity to inhibit net actin assembly downstream of Robo family receptors (Wills, 2002).
Previous data supported a simple model where Robo recruits Abl and Ena as components in the repellent pathway. In this model, Ena acts as an effector molecule to link Robo to actin assembly and Abl acts purely to antagonize and/or downregulate Robo. While this study confirms that Abl gain-of-function creates ectopic midline crossing, the additional discovery that Capt and Abl cooperate to support the repellent response and that Abl loss-of-function generates ectopic midline crossing suggests that new models are necessary (Wills, 2002).
The fact that Abl is required for midline restriction suggests that Abl plays a dual role in the Robo pathway. There are different models to explain this. As a key enzymatic component in the signaling pathway, Abl may support repellent signaling (by recruiting the necessary actin binding proteins) and also feed back on the receptor (by downregulating through phosphorylation) to adjust the sensitivity of the pathway. This model is attractive because it may explain how growth cones can adapt to different regions within a gradient of Slit. In order for a growth cone to perceive an extracellular gradient (attractive or repellent) over an extended distance, the dynamic range of the response must be continually adjusted. If the receptor system becomes saturated at any point in the gradient, the growth cone will be blind to the extracellular asymmetry at higher concentrations. Conversely, if receptor output is too low, then the signaling differential across the leading edge may be too small to detect the gradient. It has therefore been postulated that gradient guidance will require some form of adaptation to keep the signaling threshold within the appropriate dynamic range as the growth cone moves toward or away from the source. If Abl is part of the repellent response, it would also be an effective source of feedback to help match receptor sensitivity to the gradient conditions. A similar role has been postulated for MAP kinase in the Netrin signaling pathway (Wills, 2002).
The question of exactly how Abl and its signaling partners interface with the Robo receptor family is still unclear. The biochemical data suggest that Abl, Capt, and Profilin may form a large protein complex. However, the genetic interactions between Abl and robo indicate that the CC3 motif is not necessary for a functional link between Abl and Robo. This makes sense because Abl and capulet also interact with robo2, a receptor that lacks both CC2 and CC3 sequences. It is interesting that deletion of motif CC1, which is conserved in all the Drosophila Robo family members, causes a slight attenuation of the robo-Abl interaction in the assay used in this study. CC1 is also the Robo sequence phosphorylated by Abl in vitro (Wills, 2002).
Home page: The Interactive Fly © 1997 Thomas B. Brody, Ph.D.
The Interactive Fly resides on the
Society for Developmental Biology's Web server.