As an axon navigates through its environment during nervous system development, the growth cone at its tip responds to signals from many guidance cues by executing dynamic rearrangements of the actin cytoskeleton. The Rho subfamily of small GTPases—Rho, Rac and Cdc42—is critical for this modulation of the actin cytoskeleton (Hall,1992,1998; Tapon and Hall,1997). The different Rho GTPases are thought to act upon different kinds of actin structures. For example, Rho stimulates formation of focal adhesions and stress fibers (Ridley and Hall,1992), Rac promotes lamellipodial structures (Ridley et al.,1992), and Cdc42 activates filopodia (Nobes and Hall,1995).
Among the Rho GTPases, Rac has been the most enigmatic. Activation of Rac displays many effects on cell morphology, cell polarity and cell migration (Etienne-Manneville and Hall,2002; Jaffe and Hall,2005). Specifically in the growth cone, Rac plays pivotal roles in outgrowth, branching, and guidance of axons (Luo,2000b; Guan and Rao,2003). In Drosophila, Rac mutant embryos display severe axon growth defects both in central nervous system and peripheral nervous system (Hakeda-Suzuki et al.,2002). A very large number of molecules have been identified as Rac binding partners, and it is not clear which among them are the key effectors for actin rearrangement in particular contexts (Bishop and Hall,2000; Luo,2000a).
A variety of axon guidance signals seem to act through Rac GTPases (Tapon and Hall,1997; Luo,2000b). For example, in C. elegans, Rac acts downstream of UNC-40, a Netrin receptor (Gitai et al.,2003). Axonal repulsion by Slit-Robo signaling is mediated by Rac and restricted Rac function also limits Slit-Robo signaling (Hakeda-Suzuki et al.,2002; Fan et al.,2003; Yang and Bashaw,2006). The intracellular domain of Plexin B, a semaphorin receptor, binds to Rac-GTP and reduces active Rac by sequestering it from its target Pak, resulting in repulsion of axons (Vikis et al.,2000; Hu et al.,2001). Rac1 deficient cerebellar granule neurons in primary culture display reduced PAK1 phosphorylation and mislocalization of WAVE complex from the growth cone membrane (Tahirovic et al.,2010).
Like other GTPases, Rac GTPases cycle between an active GTP-bound state and an inactive GDP-bound state (Fig. 1H). Rac activity is controlled by three main classes of regulatory proteins, guanine nucleotide exchange factors (GEFs), GTPase-activating proteins (GAPs), and guanine nucleotide dissociation inhibitors (GDIs). GEFs initiate the release of GDP resulting in accumulation of GTP-bound active GTPases (Rossman et al.,2005). GAPs convert the active state to the inactive GDP-bound form (Moon et al.,2003). GDIs bind to the GDP-bound form, preventing the release of GDP and keeping Rac in the inactive state (Dovas and Couchman,2005). The overwhelming number of GEFs (∼60) and GAPs (∼70) in mammals (Etienne-Manneville and Hall,2002) suggests that each of these proteins likely has its own specific roles in different contexts.
Among the many Rac-specific GEFs and GAPs, the GEF Trio is perhaps one of the best characterized Rac regulators in axon patterning. Trio has two tandem Dbl-homology (DH-PH) GEF domains. GEF1 is a specific activator of Rac GTPase. This was shown by genetic epistasis in Drosophila, as a Rac mutation blocks the dominant effects of expressing the GEF1 domain in isolation. It was also verified by direct biochemical experiments, as Trio GEF1 activates Rac in vitro, but not Rho or Cdc42, both for Drosophila Trio (Newsome et al.,2000) and mammalian Trio (Bellanger et al.,1998). GEF1 was shown genetically to be critically required for Trio function in photoreceptor axon guidance in Drosophila, in part to regulate the activity of PAK kinase (Newsome et al.,2000). The GEF2 domain of Trio, in contrast, is a specific activator of Rho (Bellanger et al.,1998; Spencer et al.,2001) and is not required for photoreceptor axon guidance in the fly eye (Newsome et al.,2000; Vanderzalm et al.,2009). A specific axonal function of GEF1 activity was also found in C. elegans trio (UNC73; Vanderzalm et al.,2009), while GEF2 regulates pharynx and vulva musculature, synaptic neurotransmission (Steven et al.,2005) and P cell migration (Spencer et al.,2001).
The role of Trio in axon guidance has been linked to Abl tyrosine kinase (Liebl et al.,2000). Abl and Trio mutants have similar axon guidance phenotypes by themselves, and in combination they interact synergistically. Abl is one of the key molecules required in axon pathfinding (Wills et al.,2002; Forsthoefel et al.,2005; Song et al.,2008). The notion that Rac activity is required for Abl function has also been suggested in other contexts (reviewed in Hernández, 2004). Rac promotes the activities of oncogenic constitutively activated forms of Abl such as p210Bcr-Abl and v-Abl in mammalian cultured cells (Renshaw et al.,1996; Bassermann et al.,2002). Abl activates Rac in conjunction with receptor tyrosine kinase signaling in part by phosphorylation of the Ras GEF Sos-1 (Sini et al.,2004) and is also required for Rac activation following stimulation of cadherin-mediated cell–cell adhesion (Zandy et al.,2007).
We have shown previously that Abl and Trio participate in a noncanonical function of the receptor Notch in axon patterning in Drosophila (Giniger,1998; Crowner et al.,2003). In contrast to the usual Notch signaling mechanism, the function of Notch during axon guidance does not require the canonical molecular events of nuclear translocation of the intracellular domain to control target gene expression mediated by the transcription factor Su(H). Instead, Notch is present in vivo in a multiprotein complex together with Trio, and also with Disabled, another core component of Abl signaling, as shown by co-immunoprecipitation of Notch with Trio and Disabled proteins from wild-type Drosophila extracts (Le Gall et al.,2008; Song et al.,2010). This physical association of Notch with Trio and Disabled is essential for Notch-dependent control of axon growth and guidance (Le Gall et al.,2008).
Motivated by these observations, we investigated the potential involvement of small Rho GTPases in noncanonical Notch signaling during axon guidance in Drosophila embryos. Here, we first show that the Rac-specific GEF1 activity of Trio is selectively required for Trio-dependent axon patterning in embryonic motor nerves, and specifically for the interaction with Notch. Furthermore, we show a selective genetic interaction of Rac, and not Rho1 or Cdc42, with Notch, modifying its axonal function. These data support the hypothesis that Rac is a critical player in the Abl- and Trio-dependent mechanism by which Notch controls axon growth and guidance.
Trio GEF1 Activity Is Essential for Motor Axon Guidance
Motor nerve guidance in the fly embryonic nervous system provides a powerful system for quantitatively assaying the contribution of signaling proteins to axon growth and guidance in vivo. In late stage 16 embryos, subsets of motor neurons display simple and distinguishable axonal projections. Inter-segmental nerve b (ISNb) has seven motor axons that exit from the ISN root at a specific choice point to innervate ventrolateral muscles (VLM; Fig. 1A). Abl and its accessory signaling components such as Neurotactin (Nrt), Disabled (Dab), failed axon connections (Fax), and Trio are essential for proper growth and guidance of this motor nerve (Gertler et al.,1989; Hill et al.,1995; Song et al.,2010).
Like other mutations in Abl signaling, trio mutants display a specific axonal defect in ISNb, “stalling” in the middle of the target field (Fig. 1B; Table 1; Awasaki et al.,2000; Bateman et al.,2000). We found that expression of a trio transgene with a mutation inactivating the GEF1 domain (UAS-trioGEF1mu; Newsome et al.,2000) was unable to rescue the ISNb axonal phenotypes of a trio mutant. In contrast, expression of a transgene bearing the equivalent lesion in GEF2 (UAS-trioGEF2mu) rescued the axonal defects of trio as effectively as a wild-type transgene (UAS-trioWT; Fig. 1C–E; Table 1). We verified by immunostaining that all three Trio derivatives accumulated to similar levels and trafficked properly to axons (data not shown). Therefore, the activity of GEF1 is essential for the function of Trio in motor axon guidance whereas GEF2 is not. This is consistent with previous data showing that the GEF1 domain of Trio is preferentially required for sensory axon guidance in photoreceptor cells, while GEF2 is dispensable in this context (Newsome et al.,2000).
Table 1. Genetic Rescue of trio Mutant and Modification of Notch-trio Interaction by trio Transgenes in ISNb Motor Axon Guidancea
Abdominal segments A2–A7 were examined in late stage 16 embryos for quantification of ISNb phenotypes. n, total hemisegments counted. All trio transgenes were expressed by pan-neuronal elav-Gal4 driver. For rescue of trio mutant, single asterisks (*) denote statistically significant rescue by transgenes relative to trio1/trio123.4 (P < 0.005). For rescue of Notch-trio interaction, double asterisks (**) denote statistically significant modification by transgenes relative to Nts1; trio123.4/+ (P < 0.005). P values were determined by χ2 test. nsNot statistically significant. P value relative to control was more than 0.1 in each comparison. ISNb, inter-segmental nerve b.
Rescue of trio mutant by trio transgenes
Modification of N-trio interaction by trio transgenes
Nts1; UAS-trioWT; trio123.4/+
Nts1; UAS-trioGEF1mu; trio123.4/+
Nts1; UAS-trioGEF2mu; trio123.4/+
Function of Trio in Noncanonical Notch Signaling During Axonal Guidance Is Mediated by GEF1 Function
We next examined the interaction of Trio with the receptor Notch in axon growth and guidance. Inactivation of Notch at the time of axon growth selectively misroutes some motor axons, causing specific guidance defects. For example, in ISNb, Notch mutant (Nts1) axons often grow past the “choice point” at which they should exit the main ISN and enter the ventrolateral muscle domain (VLM). This results in an abnormal “bypass” innervation pattern, with few or no axon projections into the VLM (36% of total hemisegments defective, dotted area in Fig. 2B and Supp. Fig. 1, which is available online) under appropriate temperature shifting conditions (details in the Experimental Procedures section; Supp. Fig. S1). Guidance errors also occur in segmental nerve a (SNa) in Nts1 mutants (Supp. Fig. S2). The axonal action of Notch is mediated by a noncanonical signaling mechanism by which Notch locally inhibits the activity of Abl and associated cofactors (Crowner et al.,2003). Consistent with this, reduction of Abl signaling components, including Trio, suppresses the axon patterning phenotypes of Nts1. For example, heterozygosity for a loss of function trio mutation significantly restores the ability of ISNb axons to enter the VLM field in a Notch mutant genetic background (Nts1; trio123.4/+, 28%, n = 190, P < 0.05 (Fig. 2); Nts1; trio8/+, 27%, n = 236, P < 0.001). The effect of trio heterozygosity on the Notch phenotype is quantitatively modest, but it is observed with unrelated trio alleles, and is similar to the degree of suppression produced by heterozygosity for other components of the Abl signaling pathway, such as Abl and Nrt (Crowner et al.,2003). In the Discussion (below) we expand on the motivation and significance of using hypomorphic manipulations, such as heterozygous mutations, in analyzing the Notch/trio interaction.
We used the Notch/Trio interaction to further dissect the mechanism of Trio action in Notch-dependent axon guidance. We found that the suppression of Notch phenotype by a trio mutation was reverted by pan-neuronal expression of UAS-trioWT (elav-Gal4 driven), once again causing ISNb to bypass the VLM as in Nts1 (Fig. 2D,G, 86% restoration of the VLM bypass phenotype [P < 0.005], and Table 1). This suggests that neuronal trio is largely responsible for the genetic interaction of Notch with trio. Next, we examined which domains of Trio, in particular which GEF domains, contribute to the Notch-Trio interaction. Trio constructs bearing mutations that selectively inactivate either GEF1 or GEF2 domain, UAS-trioGEF1mu and UAS-trioGEF2mu, were pan-neuronally expressed in the Nts1; trio123.4/+ background. We did not detect any significant alteration of the Notch–Trio interaction with elav-GAL4-driven trioGEF1mu expression (Fig. 2E,G; 23% suppression [not significant, P = 0.14] and Table 1), while trioGEF2mu expression, in contrast, restored the bypass phenotype nearly as effectively as did UAS-trioWT (Fig. 2F,G, restore 81% of Nts1 phenotype [P < 0.005] and Table 1). Pan-neuronal overexpression of wild-type trio or GEF-inactive trio transgenes did not produce any dominant axon patterning defects in a wild-type background (data not shown). These results suggest that GEF1 activity of Trio is selectively required for the interaction with Notch in axon patterning. Consistent with this hypothesis, expression of the constitutively active Trio GEF1 domain (Newsome et al.,2000; Ferraro et al.,2007) alone mimics the Notch ISNb phenotype (48% bypass, n = 456 driven by elav-Gal4), whereas expression of Trio (GEF2) alone does not perturb ISNb patterning (less than 2% bypass, n = 228, driven by elav-Gal4).
Rac Small GTPase Is Selectively Required in Notch-Mediated Motor Axon Guidance
The observation that activity of Trio GEF1 is necessary and sufficient for the functional interaction of Notch with Trio in axon patterning led us to investigate whether Rac, the specific target for activation by GEF1, acts in this context. First, we investigated whether genetic reduction of Rac levels modifies motor axon phenotypes of Nts1, and found that heterozygosity for a Rac triple mutant, Rac1J10Rac2ΔMtlΔ/+ suppressed the axonal defects of Nts1 (Table 2, 23% of hemisegments defective, N = 132, vs. 36% for Nts1, P < 0.05). In contrast, heterozygosity for mutations of other small Rho GTPases (i.e., Cdc424 and Rho1rev220) did not alter the expressivity of ISNb bypass in Nts1 (Table 2). This result suggests that among Rho GTPases, Rac is preferentially required in the genetic pathway of Notch signaling for patterning ISNb motor axons.
Table 2. Quantification of Genetic Interactions Between Notch and Rho Small GTPases in ISNb Motor Axon Guidancea
Abdominal segments A2–A7 were examined in late stage 16 embryos for quantification of ISNb bypass phenotypes. Asterisks denote statistically significant modification by mutants and transgenes (P < 0.05). Comparison is to Nts1 for effect of mutations; comparison is to Nts1 with the appropriate GAL4 driver for transgenes. P values were determined by χ2 test. UAS-Cdc42N17 and UAS-Rho1N19 were pan-neuronally expressed by elav-Gal4 driver. Since elav-driven expression of some dominant forms of GTPases (RacN17, RacV12, Rho1V14 and Cdc42V12) and UAS-Rac1WT show massive axonal defects in most hemisegments (also described in (Kaufmann et al., 1998; Fritz and VanBerkum, 2002), for these transgenes instead we used Gal4-60, a less active GAL4 driver (Luo et al., 1994). In a completely independent set of experiments, the critical genotypes were assayed in a somewhat different temperature shift protocol and gave the same result, supporting the robustness of the genetic interaction (D. Crowner and E. Giniger, unpublished observation)
Number of hemisegments of the designated genotype that were scored.
The percentage of the hemisegments examined that exhibited ISNb bypass phenotype (described in Supp. Fig. S1).
Besides the bypass phenotype, we observed a different guidance defect, 11% of trio-like stall phenotypes.
nsNot statistically significant. P values relative to control were > 0.4 in these comparisons.
To test this hypothesis further, we used dominant-negative or constitutively active forms of Rac. Expression of a constitutively active form of Rac, RacV12 significantly increased the occurrence of ISNb bypass in Nts1 embryos (Fig. 3C; Table 2), while expression of the dominant-negative RacN17 suppressed the ISNb bypass phenotype (Fig. 3B; Table 2). In contrast, the ISNb phenotype of Nts1 was not significantly modulated by expression of dominant-negative or constitutively active forms of Cdc42 or Rho1 (Fig. 3D; Table 2). None of the GTPase transgenes produces dominant bypass phenotypes on their own under the conditions that we used, although RacN17 produces a low frequency of trio-like stall phenotypes (11%). Moreover, the effect of UAS-Rac1WT expression in Nts1 is not statistically significant (Table 2), suggesting that Rac activity, rather than amount, is critical for the genetic interaction with Notch. Taken together, our results suggest that Rac selectively modulates Notch function in motor axon guidance, while Rho1 and Cdc42 do not have strong effects in this context.
The guanine exchange factor Trio physically associates with Notch in vivo (Le Gall et al.,2008) and is a genetic component of Notch signaling in motor axon guidance (Crowner et al.,2003; Le Gall et al.,2008). In this study, we found that one of the tandem GEF domains of Trio, the Rac-specific GEF1, is essential for this genetic interaction. Consistent with this observation, the axonal phenotypes of a Notch mutant are suppressed by reducing the level of the three Rac paralogs, to a degree similar to that caused by reduction of trio. This interaction appears to be specific, because mutations of other closely related small GTPases, Rho1 and Cdc42, did not modify the axonal phenotypes of Nts1. This was further confirmed by testing the effect of dominant Rac transgenes. Expression of a dominant-negative RacN17 suppressed the ISNb phenotype of a Notch mutant, while a constitutively active Racv12 enhanced it. In contrast, introduction of other dominant transgenes such as Cdc42N17 did not alter the Notch mutant phenotypes. Thus, Rac appears to be the key Rho GTPase for control of motor axon guidance by Notch.
Our data support the hypothesis that Rac is a key component of Abl signaling in Drosophila motor axons. Notch-dependent axon patterning is executed by an alternate, “noncanonical” Notch signaling pathway defined by the Abl tyrosine kinase (Giniger,1998). Notch protein associates in vivo with the Abl cofactors, Disabled and Trio, and genetic experiments suggest that Notch antagonizes the activity of the Abl signaling pathway (Crowner et al.,2003; Le Gall et al.,2008). We find here that Rac interacts functionally with Notch in the same way as do the core Abl signaling proteins. Reduction of Rac activity suppresses Notch axonal phenotypes, just as do reduction of Abl pathway components or expression of the Abl antagonist Enabled. Enhancement of Rac activity exacerbates Notch axonal phenotypes, just as does activation of Abl signaling or mutation of Enabled. These data, therefore, are consistent with the hypothesis we proposed previously that the key role of Notch in ISNb guidance is to limit Abl-dependent adhesion of ISNb growth cones to the ISN nerve pathway (Crowner et al.,2003). By this model, excessive Abl, Rac-dependent substratum adhesion in a Notch mutant prevents ISNb growth cones from defasiculating from the ISN to enter the target muscle field. Modulation of Rac activity directly modifies this adhesion, aiding or hindering the Notch-dependent release of the growth cone from the ISN pathway at the choice point. These observations are also consistent with results from vertebrate cell culture models that suggested a crucial role for Rac in Abl-dependent signaling and cell adhesion (Zandy et al.,2007; Zandy and Pendergast,2008).
The biological function of Rac has been difficult to investigate in vivo. First, Rac performs a wide range of functions in many cells, so experimental modulation of Rac often produces complex combinations of effects. Second, the phenotypes caused by overexpression of constitutively active Rac are often similar to those produced by overexpression of a dominant-negative form, rather than being opposite, making interpretation of experimental manipulations extremely challenging (Luo et al.,1994; Luo,2000a). For example, both increase and decrease of Rac activity can cause growth cone stalling in Drosophila neurons. At the molecular level, however, this occurs for opposite reasons: extreme activation of Rac causes excessive stabilization of actin filaments, while inactivation of Rac excessively destabilizes them (Luo et al.,1994). In either event, however, the result is to block growth cone advance. These properties motivated us to introduce two technical modifications to our experiments that have been essential to obtaining clear conclusions. First, we used a sensitized genetic background in which a single molecular process was limiting for a specific axon guidance decision. This minimized the confusions normally introduced by the pleiotropic functions of Rac. We achieved this by using a precise temperature shift of a temperature-sensitive allele of Notch in a synchronized population of embryos, and assaying the turning of a single nerve comprising seven closely related axons at a precise point in their trajectory. Second, rather than using severe manipulation of Rac level or activity we used the mildest manipulations we could achieve and averaged over a large number of trials to detect quantitative modulation of an intermediate (hypomorphic) Notch phenotype by Rac. Axon growth and guidance rely on a cycle of actin dynamics. Extreme manipulations run the risk of halting that cycle altogether. We reasoned that more modest manipulations would allow us to interrogate sensitively the effect of a particular signaling molecule on the dynamics of the actin cycle. We, therefore, used heterozygous Rac mutations rather than homozygotes for investigating genetic interactions with Notch. Moreover, when expressing dominant Rac transgenes, we searched for GAL4 drivers that expressed at low, rather than high level, and that did not produce phenotypes on their own in a wild-type genetic background. These perturbations nonetheless sufficed to produce significant quantitative effects on axon phenotype in the sensitized Nts1 background. While the subtle manipulations used here necessarily produced effects that were relatively modest quantitatively, they were consistent across genotypes, such as different mutant alleles, or reduction of different genes within the Abl pathway, and they were internally consistent when comparing different kinds of perturbations, such as gain- vs. loss-of-function experiments. In contrast, use of more extreme manipulations, such as interaction with homozygous mutations in the Abl pathway, were often uninterpretable due to defects in other, unrelated developmental processes.
The data reported here suggest that Rac, acting downstream of Trio, is a major player in the noncanonical signaling pathway by which Notch controls axon growth and guidance. The key challenges now are to uncover the molecular mechanism by which Notch antagonizes Abl signaling, and to understand how and why suppression of Abl signaling by Notch promotes proper growth and guidance of Notch-dependent axons.
Fly stocks were raised on standard Drosophila media at room temperature (23–25°C) except for Notch mutant, Nts1 in 18°C. We obtained fly stocks as follows: trio123.4-E. Leibl (Dennison University, OH); Gal4-60-G. Technau (University of Mainz, Mainz, Germany); Rho1rev220, S. Parker (Fred Hutchinson Cancer Research Center, WA); elav-Gal4, Y.N. Jan (UCSF, CA); Rac1J10 Rac2ΔMtl, UAS-RacN17, UAS-RacV12, UAS-Rac1, UAS-Cdc42V12, UAS-Cdc42N17 (L. Luo, Stanford University, CA); Cdc424 (R. Fehon, Duke University, NC); trio1,UAS-trioWT, UAS-Rho1N19, UAS-Rho1V14 (the Bloomington Drosophila Stock Center, Bloomington, IN). To examine the specificity of GEF activity of Trio, we generated transgenic lines of GEF1-inactive (UAS-trioGEF1mu) and GEF2-inactive (UAS-trioGEF2mu; mutant trio constructs were kindly provided by B.J. Dickson). Chromosome balancer containing β-galactosidase (TM6B-T8-LacZ and CyOact-LacZ) were used in all genetic experiments.
Sample Collection and Preparation
Egg collection and fixation was carried out as previously described (Crowner et al.,2003) with a slight modification of temperature shift condition; 0–3 hr Nts1 embryos were placed at 18°C for 13 hr and then 32°C for 6 hr. In this late-shifting condition, a significant portion of motor axons display abnormal patterning including ISNb phenotype (Supp. Fig. S1) without affecting relevant cell fates, including number and location of neurons and glia (confirmed by marker staining, data not shown) and morphogenesis of muscles. For scoring SNa phenotypes, embryos were placed at 18°C for 14 hr and then 32°C for 5.5 hr.
Collected embryos were stained with antibodies by standard methods (Bodmer, 1987; Bier, 1989). Anti-Sxl (M114, Developmental Studies Hybridoma Bank, Iowa City, IA) was used at 1:50 for segregating hemizygote for X chromosome, negative for Sxl expression. Anti-fasciclin II (1D4, Developmental Studies Hybridoma Bank) antibody was used at 1:150 for labeling motor nerves. Rabbit anti β-gal (1:1,000, Cappel, preabsorbed before use) was used for sorting negative embryos out for further analysis.
ISNb was scored in hemisegments A2–A7 by staining with anti-fasciclin II, as described previously (Song et al.,2010). Data for a given genotype were pooled and significance vs. control was assessed by χ2 test. Average variation between duplicate trials was less than 15% of the mean value for a given condition (Table 2).
We thank K. Zinn, L. Belluscio and the members of the Giniger Lab for comments on the manuscript, D. Crowner and Z. Liu for technical assistance, the Bloomington Drosophila Stock Center for fly lines, and Developmental Studies Hybridoma Bank for antibody reagents. E.G. was funded in part by the Intramural Research Program of the NIH (NINDS: Z01NS003013).