|Home | About | Journals | Submit | Contact Us | Français|
Notch is required for many aspects of cell fate specification and morphogenesis during development, including neurogenesis and axon guidance. We here provide genetic and biochemical evidence that Notch directs axon growth and guidance in Drosophila via a “non-canonical”, ie non-Su(H)-mediated, signaling pathway, characterized by association with the adaptor protein, Disabled, and Trio, an accessory factor of the Abl tyrosine kinase. We find that forms of Notch lacking the binding sites for its canonical effector, Su(H), are nearly inactive for the cell fate function of the receptor, but largely or fully active in axon patterning. Conversely, deletion from Notch of the binding site for Disabled impairs its action in axon patterning without disturbing cell fate control. Finally, we show by co-immunoprecipitation that Notch protein is physically associated in vivo with both Disabled and Trio. Together, these data provide evidence for an alternate Notch signaling pathway that mediates a postmitotic, morphogenetic function of the receptor.
We have come to appreciate that a small handful of ubiquitous, highly conserved signal transduction pathways together account for a remarkable fraction of the patterning of a developing animal. In different tissues, Hedgehogs, Wnts, TGFβs, receptor tyrosine kinases and a few other key signaling modules specify cell identities, set developmental boundaries, direct cell migration and shape tissue morphogenesis (Gerhart and Kirschner, 1997; Ptashne and Gann, 2002). In each case, however, it remains mysterious how a single signal can produce such varied effects. In the case of Wnt signaling, some of the diversity of biological readouts seems to arise from the existence of at least three different Wnt signaling pathways, the “canonical” β-catenin pathway and at least two “non-canonical” signaling mechanisms (Boutros et al., 1998; Kuhl et al., 2000; Pandur et al., 2002; Yoshikawa et al., 2003). Presumably, this diversity of signaling pathways is used, in various combinations, to elicit different cellular outcomes in response to an input Wnt signal. Multiple signaling pathways have also been described downstream of other receptor families, such as the receptor tyrosine kinases (Fantl et al., 1992).
The receptor Notch, together with its ligands, Delta and Serrate, define another of these common, multipotent developmental signaling pathways (Artavanis-Tsakonis et al., 1999; Frisen and Lendahl, 2001). First studied in Drosophila, Notch orthologs have been found in nearly all metazoan phyla, where they define the boundaries of developmental compartments, distinguish the developmental potentials of sibling cells and limit the segregation of differentiated cells from among fields of equipotent progenitors. A conserved signaling mechanism has been described for Notch proteins, whereby ligand activation causes Notch to be cleaved proteolytically at the inner edge of the plasma membrane, releasing an intracellular fragment that transits to the nucleus to form a transcription control complex in association with at least two cofactors, the DNA-binding protein Su(H) (a member of the CSL family of proteins), and the transcriptional co-activator Mastermind (Mam)(Artavanis-Tsakonis et al., 1995; Hansson et al., 2004; Jeffries et al., 2002; Petcherski and Kimble, 2000).
Over the past several years, however, experiments in a variety of vertebrate and invertebrate systems have hinted at the existence of a “non-canonical”, Su(H)-independent, signaling pathway for Notch (Brennan and Gardner, 2002; Ordentlich et al., 1998; Ramain et al., 2001; Shawber et al., 1996; Wang et al., 1997; Zecchini et al., 1999). In Drosophila, for example, it seems that a Su(H)-independent Notch activity regulates the morphogenetic movements of dorsal closure (Zecchini et al., 1999), while in vertebrates, CSL-independent Notch signaling appears to block muscle differentiation of myoblasts (Shawber et al., 1996), inhibit transcriptional activation by E-proteins (E47) in B-cells (Ordentlich et al., 1998) and promote survival of neural stem cells and embryonic stem cells (Androutsellis-Theotokis et al., 2006). While these studies have been suggestive, however, it has been difficult to test thoroughly the notion of a non-canonical Notch signaling pathway since in most cases the molecular properties of the postulated alternate pathway have not been clearly defined.
Non-canonical Notch signaling has also been invoked in axon guidance in Drosophila (Crowner et al., 2003; Giniger, 1998), and we have suggested that the Abl tyrosine kinase and its associated signaling pathway provides a good candidate for an alternate Notch signaling mechanism. This hypothesis was based upon two major findings, that Notch interacts genetically with gain- and loss-of-function manipulations of abl and its cofactors to direct the growth of particular axons, and that a protein interaction domain of one postulated Abl accessory factor, the adaptor protein Disabled (Dab), can bind directly to Notch in vitro. A link between Notch and Abl is of interest not just because of its implications for Notch, but also for the light it might shed on Abl signaling. Abl is a cytoplasmic tyrosine kinase that has been studied for more than two decades as one of the first cellular genes clearly implicated in the etiology of a common human cancer (Druker et al., 2001; Fainstein et al., 1987; Goff et al., 1980). Moreover, a central role for Abl in axon patterning has been solidly established by extensive genetic experiments, particularly in Drosophila (Hoffmann, 1991). Here, Abl contributes to the growth and guidance of many, if not most, developing axons, apparently by locally controlling actin organization and dynamics (Baum and Perrimon, 2001; Bear et al., 2000; Gertler et al., 1989; Grevengoed et al., 2001; Luo, 2000; Wills et al., 1999).
Abl performs its functions in cooperation with a constellation of conserved accessory proteins. Three cooperating factors were originally identified in Drosophila as genetic loci that enhance the phenotype of an abl mutant. These were trio, which encodes a guanine nucleotide exchange factor (GEF) for Rho GTPases, fax, encoding a protein found in axon bundles, and neurotactin (nrt), encoding a cell-surface adhesion protein (Gertler et al., 1993; Hill et al., 1995; Hortsch et al., 1990; Liebl et al., 2000; Liebl et al., 2003). An additional putative cooperating factor, disabled (dab), was originally identified as a multicopy suppressor of both fax abl and abl nrt mutant combinations (Gertler et al., 1993; Liebl et al., 2003). In addition to these cooperating factors, Abl works in concert with a specific antagonist, the signaling protein Enabled (Ena (Gertler et al., 1995)). Despite intensive study, however, to date it has been difficult to discern the detailed mechanism by which Abl pathway proteins guide axon growth. For example, there have been only a handful of cases where it has been possible to demonstrate physical association of Abl pathway effector proteins with growth cone receptors in vivo, and to correlate those associations with specific axon patterning decisions (Bashaw et al., 2000; Forsthoefel et al., 2005).
We present here genetic, molecular and biochemical evidence for a non-canonical, ie non-Su(H) mediated, Notch signaling pathway employing Abl signaling components to promote axon growth and guidance in Drosophila. We prepare Notch derivatives that are inactive for the Su(H)-dependent cell fate function of Notch but find that they nonetheless are largely or fully active for promoting Notch-dependent axon patterning, while conversely, removing the binding site on Notch for Disabled impairs axon patterning without affecting cell fate control. We also show that Notch co-immunoprecipitates from wild type fly lysates in association with Disabled, and with the Abl pathway effector, Trio. Together, these data suggest a molecular basis for the genetic interaction of Notch and Abl that promotes embryonic axon patterning in Drosophila, as well as providing reagents and diagnostic properties that can be used to identify the contribution of this pathway at other times and places in development.
Previous experiments have shown that inactivating Notch with a temperature-sensitive mutation (Notchts1) shortly before CNS longitudinal axon tracts are pioneered in the developing fly embryo yields a ‘disconnected’ phenotype in which axons fail to grow between segments of the ventral nerve cord, and those axons can be restored by expressing a wild type Notch transgene in postmitotic neurons, (Fig 1 B-H, quantified in Fig 1N; Giniger, 1998; Giniger et al., 1993). We therefore expressed a series of modified Notch transgenes in the Notchts1background to map domains that are required for axonal function of the protein. The same derivatives were also tested for their ability to perform “lateral inhibition”, the classic Notch cell fate function that limits the number of neurons in the embryo, and whose absence produces the severe neural hyperplasia characteristic of strong Notch alleles (known as the “neurogenic phenotype” (Lehmann et al., 1983), Fig 1 I-M; (Le Gall and Giniger, 2004)).
Deletion of the binding region for the ligands Delta and Serrate (EGF repeats 10-12; Lieber et al., 1993; Rebay et al., 1991) from the Notch extracellular domain completely ablated the ability of a Notch transgene to restore axon patterning (Figure 1G; quantification in panel N), consistent with prior experiments showing that a Delta mutation produces the same axonal defects as does a Notch mutation (Crowner et al., 2003; Giniger et al., 1993). Intracellular sequences of Notch were also essential for axon patterning as a transgene deleted for nearly the entire intracellular domain (NotchECN (Jacobsen et al., 1998)) was devoid of axonal activity (Fig 1H).
Deletion of a portion of the intracellular domain comprising the ankyrin/cdc10 repeats and the C-terminal portion of the Ram domain (NotchΔcdc10) reduced axonal activity of Notch significantly, however, to our surprise substantial axonal activity remained (Fig 1E, N; comparison to Notchts gives P 0.001; χ2). This is in contrast to the complete failure of this protein to suppress the neural hyperplasia of a strong Notch mutant (Fig 1L, similar to the Notch null embryo in 1J), or to activate expression of canonical Notch target genes (reviewed in Le Gall and Giniger, 2004). Remarkably, moreover, a Notch derivative truncated at the C-terminal end of the ankyrin repeat region, NotchΔ2155 (Wesley and Saez, 2000), was nearly wild type for Notch axonal function, even though it, too, was completely inactive for suppressing the Notch “neurogenic” phenotype (Fig 1 F, M) and has been shown previously to be incapable of activating expression of the Notch target gene, E(spl) (Wesley and Saez, 2000). These data begin to suggest that the sequence requirements for Notch function in axon patterning may be different from those required for the canonical cell fate functions of the protein.
Abl tyrosine kinase has been implicated genetically in the mechanism by which Notch controls axon patterning, and one postulated Abl accessory protein, the adaptor protein Disabled, binds in vitro to a small region in the juxtamembrane portion of the Notch intracellular domain (Giniger, 1998; Le Gall and Giniger, 2004). We therefore wondered whether the axon patterning activity of Notch might depend, in part, on the Disabled binding site, which is still present in both NotchΔcdc10 and NotchΔ2155. As above, modified Notch transgenes were assayed for their ability to restore axon patterning in Notchts1embryos and to suppress neural hyperplasia in embryos hemizygous for the strong allele, Notch55e11 (Giniger, 1998; Le Gall and Giniger, 2004).
Deletion of Su(H) binding sites of Notch had only limited effect on axon patterning despite ablating almost completely the ability of Notch to limit neurogenesis, while deletion of the Disabled site impaired axonal function with little or no effect on neurogenesis (Figure 2). A derivative of full-length Notch deleted for the Disabled binding site, NotchΔRamA, was impaired by about 30% relative to the wild type protein in its ability to restore longitudinal growth of CNS axons when expressed in Notchts1 (Fig 2 B, quantified in I; P 0.001 (χ2)). The minimal deletion we have so far identified that prevents Disabled binding also deletes the amino-terminal Su(H) binding site, thus, NotchΔRamA has a slight deficit in lateral inhibition (Fig 2F; quantified in J). However, a Notch protein bearing a missense mutation that inactivates just that Su(H) site, (NotchRam* (Le Gall and Giniger, 2004)) has an identical deficit in neurogenesis (Fig 2 G; P > 0.5, ANOVA) but is nearly wild type for axon patterning (Fig 2 C, I), suggesting that the axonal phenotype of NotchΔRamA is not due to the absence of this Su(H) binding site. Indeed, even a Notch derivative lacking all three Su(H) binding sites (NotchRam*Δ3) retains significantly more axon patterning activity than does NotchΔRamA (Fig 2D, p < 0.01; χ2), even though Notch Ram*Δ3 is almost completely inactive in suppressing neural hyperplasia (Fig 2H; compare with the level of neural hyperplasia in dorsal portions of the Notch mutant embryo in Fig 1J, and see also (Le Gall and Giniger, 2004)), whereas NotchΔRamA is only mildly impaired. (Note that the efficacy of the NotchRam*, NotchΔRamA and Notch Ram*Δ3 mutations for blocking Su(H) binding to Notch has been documented previously (Le Gall and Giniger, 2004).) We wished to be certain that the differential rescue of axonal versus cell fate phenotypes by these Notch derivatives reflected the differing requirements of the two biological processes and not peculiar properties of the Notchts1 protein. In separate experiments, therefore, we also induced neurogenic defects by early temperature shifts of Notchts1 and confirmed that they displayed the same pattern of rescue by various Notch derivatives as did the neurogenic defects of Notch55e11 (data not shown). We therefore infer (1) that the ability of Notch to control axon patterning cannot be accounted for by its binding sites for Su(H), (2) that the Disabled site, and not the Su(H) site, within the ΔA deletion is likely to be responsible for the axonal deficit of NotchΔRamA protein, and (3) that the Disabled site does not contribute substantially to the neurogenic function of Notch.
Similar results were obtained from assay of a second Notch-dependent axon guidance decision. An appropriate temperature shift of Notchts embryos blocks defasciculation of ISNb motor axons from the intersegmental nerve (ISN) at the first peripheral choice point ((Crowner et al., 2003), shown schematically in Fig 3A). Since this is a relatively late event in embryogenesis, the temperature shift does not perturb CNS development, simplifying the phenotypic analysis and facilitating precise quantification. A Notch transgene lacking just the Ram Su(H) site or all three Su(H) sites rescued axon patterning as well as did a wild type Notch transgene (Fig 3B; 8% of hemisegments showing residual axonal defects in embryos rescued with NotchRam* and 7% with Notch Ram*Δ3, vs 6% for wild type Notch; difference not significant). In contrast, the transgene lacking the Disabled binding site is significantly impaired in axonal activity, with rescue reduced by ~40% compared to wild type (p < 0.01; ANOVA).
As an alternate way to assess the role of canonical Notch signaling in axon patterning we examined expression of the known Notch target gene, E(spl)mδ, in cells that are executing Notch-dependent guidance processes. Previous experiments have shown that neuronal Notch activity is required for extension of the axons of the longitudinal pioneers dMP2 and vMP2, and for defasciculation of the ISNb pioneer RP3 (Giniger, 1998; Crowner et al., 2003). Examination of an E(spl)mδ-lacZ reporter gene (Cooper and Bray, 1999), however, fails to reveal any evidence for lacZ expression in these pioneer neurons, or in surrounding neurons, at the time these Notch-dependent axon guidance events are occurring (Figure 4). These data should be interpreted cautiously, of course, since even the most general Notch target genes are not used in all Notch-dependent processes, however, this observation is consistent with the idea that activation of Notch signaling for guidance of CNS longitudinal pioneers and ISNb motoneurons may not be associated with activation of the canonical effector pathway.
Data above suggest that the Disabled binding site on Notch is required for Notch-dependent axon patterning, a process mediated by the genetic interaction of Notch with genes of the Abl signaling pathway. The role of Disabled in Abl signaling, however, remains controversial due to the lack of any disabled mutants with which to assess the loss of function phenotype of the gene (Liebl et al., 2003). To further test the role of Disabled in Notch/Abl-dependent axon patterning we therefore assayed the effect of Disabled overexpression on defasciculation of ISNb (Fig 5). In a wild type genetic background, overexpression of Disabled had no effect on ISNb defasciculation at the first choice point (bypass of the choice point observed in 3 ± 1% of hemisegments, the same as in wild type embryos subjected to the temperature shift protocol (Crowner et al., 2003; Wills et al., 1999)). However, Disabled overexpression substantially enhanced the Notch mutant phenotype (42 ± 0.3% bypass in Notchts; elav-GAL4; UAS-Disabled, vs 32 ± 0.7% in the Notchts internal control embryos; P 0.01, ANOVA). This synergistic interaction of Notch with Disabled mimics the interaction of Notch with other manipulations that enhance signaling in the Abl pathway, including overexpression of Abl or heterozygous mutation of the Abl antagonist, enabled (Crowner et al., 2003), consistent with the hypothesis that Disabled modulates Abl pathway signaling in vivo.
In support of the genetic data described above, biochemical experiments showed that Notch co-immunoprecipitates from wild type Drosophila lysates with Disabled, and also with the Abl accessory protein, Trio. In the case of Disabled, Western blotting of anti-Disabled immunoprecipitates revealed the presence of full-length Notch (Fig 6). Four lines of evidence suggest that the association of Notch with Disabled is authentic and specific. First, three different anti-Disabled antibodies co-precipitated Notch. These included two independent polyclonal antibodies (one is shown in Fig 6A, lane 4, middle panel), as well as a monoclonal antibody (Fig 6B, lane 4; characterization of anti-Disabled antibodies is documented in Supplementary Figure S-1). Second, association was detected in two different wild type fly lysates, a total embryo lysate (Figure 6B) and a lysate made from adult heads (Fig 6A). Third, immunoprecipitations performed in parallel with two irrelevant mouse IgGs (anti-Islet, or normal mouse IgG) did not co-precipitate detectable Notch protein (Fig 6A, lane 5, middle panel; Fig 6B, lane 3; Fig 6C, lane 3; data not shown). Fourth, immunoprecipitation of Disabled from extracts of Notch-overexpressing animals co-precipitated more Notch protein than did immunoprecipitation from wild type extracts done in parallel. We have not been able to test for the presence of Disabled in anti-Notch immunoprecipitates since the current anti-Disabled antibodies, while specific, are not sufficiently sensitive in Western blots (Fig 6A, bottom panel).
Multiple lines of evidence also demonstrate association of Notch with Trio in wild type fly extracts. Immunoprecipitations with anti-Notch caused co-precipitation of Trio, again, both from adult head lysates (Figure 6A, lane 3, top panel) and from wild type embryo lysates (not shown). We also reversed the experiment and found that immunoprecipitation of Trio caused co-precipitation of full-length Notch (Fig 6C, lane 4). Finally, side-by-side co-immunoprecipitations of extracts from wild type and Notch-overexpressing animals detected more Notch co-precipitated with anti-Trio in the overexpressing lysates (Suppl Info, Fig S-2, compare lanes 3, 4), as well as detecting more Trio in anti-Notch immunoprecipitates from these extracts (not shown). Together, these data show that Notch is found in complexes with both Disabled and Trio in extract of wild type flies. Our data do not prove whether Notch associates with Disabled and Trio simultaneously or whether these two proteins form separate complexes with Notch, however, we note that immunoprecipitations with any of three anti-Disabled antibodies reliably co-precipitated Trio, both from adult (Fig 6A, lane 4, top panel) and embryo extracts, suggesting that Trio and Disabled are indeed found in common complexes in vivo.
Previous studies have led us to speculate that the Abl tyrosine kinase and its associated accessory factors might define an alternate, ‘non-canonical’, Su(H)-independent signaling pathway for the receptor Notch (Crowner et al., 2003; Giniger, 1998). The data reported here provide strong support for this hypothesis. In extracts of wild type Drosophila, Notch is associated with Disabled and Trio, two proteins that have been associated with the action of Abl tyrosine kinase. The functions of Notch in axon growth and guidance are likely to be executed by these complexes of Notch with Disabled and Trio, and not by its association with Su(H), since deletion of the Disabled binding site from Notch significantly impairs the axon patterning function of the receptor, whereas the Su(H) binding sites are largely dispensable for this process. Moreover, we describe two other Notch derivatives that are still capable of executing the axon patterning functions of the receptor despite being completely inactive for specifying Notch-dependent cell fates. Taken together with previous data demonstrating that the genetic interaction of Notch with multiple Abl pathway components is required specifically for Notch-dependent axon patterning, these data provide a molecular picture of a Notch signaling machinery that is distinct from the well-established mechanism by which a proteolytic fragment of Notch enters the nucleus to directly control transcription of Su(H)-dependent target genes.
The key genetic data in favor of our hypothesis stem from the targeted construction of Notch derivatives that preferentially impair either the canonical, cell fate function of the receptor or its Abl-dependent axon patterning function, respectively. Deletion of the Su(H) binding sites from Notch progressively and dramatically reduces the ability of the receptor to limit neurogenesis, but has only limited effect on growth of CNS longitudinal axons, and no detectable effect on Notch-dependent defasciculation of ISNb motor axons. In contrast, deletion of the Disabled binding site substantially reduces the axon patterning activity of Notch (30-40%, depending on the assay), while having no effect on cell fate function beyond what can be accounted for by the known Su(H) binding site within the deletion. The properties of these complementary Notch mutants argue for the action of a qualitatively different Notch mechanism in axon patterning. Further supporting this hypothesis is the observation that Notch co-precipitates from wildtype fly extracts with two cytoplasmic signaling proteins, the Abl cofactor, Trio and the adaptor protein, Disabled, potentially providing a molecular machinery to account for our phenotypic data. In principle, a good way to further test the basis of the Notch axonal phenotype would be to examine a disabled mutant, but unfortunately no such mutants are currently available (Liebl et al., 2003). The phenotype of a trio mutant, on the other hand, is consistent with the results documented here (Bateman et al., 2000; Hakeda-Suzuki et al., 2002; Liebl et al., 2000). The zygotic mutant phenotypes of trio are somewhat subtle, evidently because of persistence of maternally-provided trio RNA and protein, but they include defects in some of the CNS longitudinal axons that are affected in Notchts embryos, as well as defects in ISNb motor axon guidance, while trio has not been reported to produce any Notch-like defects in cell fate (Bateman et al., 2000; Liebl et al., 2000).
While deletion of the Disabled binding region of Notch clearly reduces the axonal activity of the protein, substantial residual activity still remains. In the case of Su(H), residual activity of a Notch mutant lacking all in vitro Su(H) binding sites can be traced to an association of Su(H) with the Notch ankyrin repeats that requires the cofactor, Mastermind (Nam et al., 2006; Wilson and Kovall, 2006; Wu et al., 2000). By analogy, perhaps Disabled can also associate with Notch via a second site that requires a cofactor present in vivo. Consistent with this idea, preliminary biochemical experiments hint that the ΔRamA mutation does not wholly ablate recruitment of Disabled and Trio in vivo (MLG and EG, unpublished observations), though current reagents do not allow us to assess this rigorously. If so, the ankyrin/cdc10 repeat region would be a plausible candidate for a secondary site of association. Experiments in Fig 1 show that the ankyrin repeats, together with the C-terminal portion of the Ram region, contribute substantially to Notch-dependent axon patterning. Since our experiments clearly show that the canonical Notch signaling pathway is dispensable for axonal function, the axonal requirement for this portion of the protein cannot be traced to its function in canonical signaling, and a contribution to formation or activity of Notch/Abl pathway complexes would offer the simplest explanation.
To date, the role of Disabled in the Abl signaling pathway has been difficult to establish due to the lack of loss-of-function Disabled mutant alleles. The placement of Disabled in the Abl pathway was based initially largely upon the observation that modest overexpression of Disabled suppressed both the embryonic lethality and morphological defects produced by genetic interactions of Abl with its accessory genes Nrt and Fax (Gertler et al., 1993; Liebl et al., 2003). Our data now show that deletion of the Disabled binding region of Notch specifically impairs a Notch function, axon patterning, that depends on the interaction of Notch with multiple Abl pathway components, (Figs (Figs2,2, ,3;3; (Crowner et al., 2003; Giniger, 1998)). Moreover, GAL4-driven overexpression of Disabled modifies the Notch ISNb phenotype in the same way as do other treatment that enhance Abl signaling, such as overexpression of Abl or reduction of enabled (Fig 5; Crowner et al., 2003). Thus, these data provide further support for association of Disabled with the Abl signaling pathway, though a definitive demonstration awaits the generation and characterization of a disabled mutation.
The presence of Trio in Notch complexes suggests that Rho family GTPases, particularly Rho and Rac, are good candidates for a downstream readout of Notch/Abl signaling (Hakeda-Suzuki et al., 2002; Luo, 2000; Newsome et al., 2000). Consistent with this we have observed dominant genetic interactions of Notch with mutations in the three Drosophila Rac genes, but not, for example, with Cdc42 (EG, unpublished observations). Such a readout would make sense in the context of the effects of Notch on growth cone guidance and would be consistent with previous studies of Drosophila Trio (Hakeda-Suzuki et al., 2002; Newsome et al., 2000) and mammalian Abl (Renshaw et al., 1996). It is interesting to note that identification of Rac as an effector of Notch/Abl signaling might suggest the possibility of Notch/Abl signaling having a non-Su(H) nuclear component in some developmental contexts in addition to its cytoskeletal targets (Hall, 2005; Lim et al., 1996). Rho family GTPases typically have multiple downstream targets, including nuclear gene regulation in addition to cytoskeletal structure and dynamics (Luo, 2000).
The key step in canonical Notch signaling is the proteolytic cleavage of the receptor by γ-secretase to release the active, intracellular moiety of the molecule, NICD. Does γ-secretase also play a role in Notch/Abl signaling? Since we find Disabled and Trio associated with full-length Notch prior to cleavage, and since Notch/Abl signaling in the growth cone presumably targets the cortical actin cytoskeleton, one possibility is that γ-secretase cleavage terminates the Notch/Abl signal by separating the receptor-bound complex from membrane-tethered components of the pathway such as Abl kinase and Rho GTPases. Alternatively, in contexts such as ISNb, perhaps displacement of Disabled and Trio away from the membrane is part of the mechanism by which Notch antagonizes Abl pathway activity. Moreover, while proteolytic activity is the most apparent function of γ-secretase there are have been suggestions that the complex may also have a separate function in Notch trafficking, aside from cleavage. If so, this activity could modulate Notch/Abl signaling independent of any role for protease cleavage. Clearly, additional experiments will be necessary to assess the various possible models.
Is the interaction of Notch with Abl pathway proteins limited to just a few Drosophila growth cones, or is it of more general biological significance? Our ability to detect Notch complexes with Disabled and Trio in extract of whole embryos argues for the latter, as does the strong phylogenetic conservation of all the components of the pathway. Good candidates for potential Notch/Abl-dependent processes are provided by those developmental contexts in which non-Su(H) Notch signaling has been proposed previously. In Drosophila, these include organization of actin structure at the D/V boundary of the developing wing (Major and Irvine, 2005); in mammals, they include myogenesis and B-cell development, as well as a ligand-stimulated cytoplasmic signaling process of Notch that is essential for the survival of mouse neural stem cells and human embryonic stem cells (Androutsellis-Theotokis et al., 2006; Ordentlich et al., 1998; Shawber et al., 1996).
Axon guidance and classic lateral inhibition seem to represent limit cases in which the Notch signal is largely transduced selectively through either the Abl pathway or the Su(H) pathway, respectively. It seems likely, however, that in each case both pathways make some contribution to Notch function: deletion of Su(H) binding sites does have some deleterious effect on growth of CNS longitudinal axons (Fig 2), while mutation of Abl and Nrt cause small but reproducible decreases in the efficacy of the classic Notch function that discriminates the identities of sibling cells (Giniger, 1998; EG, unpublished observations). Perhaps two parallel Notch signals, one through the canonical Su(H) pathway and the other mediated by the Notch/Abl interaction, can be used in concert to provide a richer nuclear readout, or to coordinate nuclear gene regulation with cortical properties such as cytoskeletal structure and cell adhesion. It will be of great interest to determine whether some classic functions of Notch, such as dendritic patterning (Berezovska et al., 1999; Redmond et al., 2000; Sestan et al., 1999) or oncogenesis (Raafat et al., 2007), reflect more balanced contributions both from canonical Notch signaling and from the Notch/Abl pathway.
Rescue of Notch axonal and cell fate defects by UAS-Notch transgenes was performed as described previously (Crowner et al., 2003; Giniger, 1998; Le Gall and Giniger, 2004), with the modification that for the experiment of Fig 2, assay of CNS axon extension employed the following temperature-shift protocol to improve the precision of quantification: embryos were collected 4 hours at 18°; aged 7 hours, 18 °, shifted 7h 45′ (32 °), then fixed. To induce neurogenic defects in Notchts1, embryos were collected 6 hours (18 °), shifted 10 hours (32 °) and then fixed. UAS-NotchΔcdc10 and UAS-NotchΔEGF10-12 flies were from A. Martinez-Arias (Zecchini et al., 1999), UAS-NotchECN was from M. Muskavitch (Jacobsen et al., 1998). The UAS-NotchΔ2155 transgene was obtained from C. Wesley (Wesley and Saez, 2000) and transformant lines were generated by injection into w1118. UAS-Disabled was from D. Van Vactor (Wills et al., 1999). All other mutants and UAS transgenes have been described previously (Giniger, 1998; Le Gall and Giniger, 2004) except UAS-NotchΔRamA. For this construct, a deletion of Notch codons T1766-R1801 was generated by PCR and reconstructed into full-length UAS-Notch as described for other juxtamembrane derivatives (Le Gall and Giniger, 2004).
Embryos for immunocytochemistry were collected, fixed and stained by standard methods (Bodmer and Jan, 1987). Detection was performed using biotinylated secondary antibodies and the Vectastain Elite tertiary reagent (Vector Labs), with DAB development. Embryos were examined either as whole-mounts in JB4 embedding medium or as filet preps in 80% glycerol. Fluorescence microscopy was performed on filleted samples mounted in FluoroGuard (BioRad), using a Zeiss AxioImager and deconvolution of the resulting image stacks.
Antibodies, their sources and the dilutions used were as follows. For embryo immunostaining, anti-Fasciclin 2 (mAb 1D4, 1:150), anti-Sxl (mAb M114; 1:50), mAb 22C10 (1:25) and rat anti-Elav (mAb 7E8A10; 1:20) were obtained from the Developmental Studies Hybridoma Bank; rabbit anti-β-galactosidase (1:10,000) was from Cappel. For biochemical experiments, immunoprecipitations were performed using 1-10 μg of specific antibody per sample, and Western blotting was performed using 10-50 μg specific antibody per 10 ml probe solution. Anti-Notch (C17.9C6), anti-islet (mAb40.3A4). and anti-β-tubulin (E7) were from the DSHB. Rabbit anti-Trio B6 and mouse anti-Trio (mAb 9A4) were gifts from C. Hama (Awasaki et al., 2000). A high-concentration stock of purified mAb9A4 was generated by the Antibody Production Facility of the Fred Hutchinson Cancer Research Center, using cells provided by Dr. Hama. Non-immune mouse IgG, rabbit-anti-mouse IgG, and all secondary antibodies were from Jackson ImmunoResearch.
Antibodies against Drosophila Disabled protein were prepared as follows. The coding sequence for a fragment of Disabled corresponding to the amino-terminal 2/3 of the protein (from the PTB domain up to an EcoRI site at nt5868 in the numbering of Gertler, et al., 1993) was subcloned in frame into pGEX2T to generate a GST fusion construct, and into pRSET A to generate a His-tagged construct. After transformation of expression constructs into E. coli, tagged protein was grown up at room temperature and purified on glutathione beads or nickel beads, respectively, by standard methods. Several polyclonal anti-Disabled antisera were prepared by Dr. Liz Wayner of the FHCRC Antibody Development Facility by injection of either fusion protein into mice, using standard protocols. Two mice that gave ELISA-positive bleeds after injection with the His-tagged immunogen were further used to generate hybridomas; these were screened by ELISA, Western blotting, immunoprecipitation and whole mount embryo staining. mAb P4D11 was found to immunoprecipitate a set of species with the published characteristics of Disabled from extracts of wild type Drosophila; mAb P6E11 recognized on a Western blot a family of high-molecular weight species of a size consistent with Disabled, and which were absent from extract of embryos deleted for the Disabled locus (Supplementary Figure S-1). Both antibodies also labeled (albeit with some nonspecific background, especially for mAb P4D11) a pattern of tissues in whole mount embryos consistent with the Disabled expression pattern and absent from dab-deficient embryos. A high concentration stock of mAb P6E11 ascites was prepared by Maine Biotechnology.
Phenotypes were quantified as follows:
For CNS longitudinal axons, Fasciclin 2-labelled embryos were examined either as filet preps, or in appropriately oriented whole mounts, and the number of thoracic and abdominal hemisegments in which there was a complete or nearly complete disconnection between successive segments was counted and expressed as a percentage of the total. A “nearly complete” disconnection was defined as one in which the residual connection between hemisegments consisted of a single Fasciclin 2-positive axon bundle with less then ~1/5 the staining intensity of a wildtype fascicle. In a single experiment, either filet preps or whole mounts were quantified; quantitative data were not compared between different kinds of preparations. Note that slightly different temperature-shift protocols were employed for the experiments of Figure 1 vs Figure 2 (see above) so expressivity per hemisegment should not be compared between these two experiments.
ISNb defasciculation was assayed in abdominal segments A2-A7 in filet preps of Fasciclin 2-labelled embryos (Crowner, et al., 2003). A hemisegment was scored as an aberrant “bypass” hemisegment if at least one Fasciclin 2-positive ISNb axon failed to defasciculate from the ISN and project internally at the wild type choice point, immediately ventral to muscle 28, and instead projected external to muscle 28. The fraction of bypass hemisegments was scored in greater than 100 hemisegments of each of at least three independent experiments. Reproducibility (SEM) was ± 3% for this assay.
Notch neurogenic phenotypes were quantified by counting the number of Elav-immunoreactive nuclei in the dorsal cluster of sensory neurons in hemisegments A1-A7. SEM was ≤ 1.0 neurons per dorsal cluster for all genotypes
Lysis buffer for both embryo lysates and head lysates was 25mM Tris (pH 7.5), 100 mM NaCl, 0.5 mM DTT, 0.5% NP-40 (v:v), 10% glycerol (v:v), 1mM PMSF + a 1:100 dilution of a cocktail containing 1 mg/ml leupeptin, 1 mg/ml pepstatin, 2 mg/ml aprotinin and 10 mg/ml benzamidine (prepared in DMSO). For whole embryo lysates, embryos were collected on grape juice plates 18-24 hours at 25°. Embryos were harvested with 0.7% NaCl/0.3% Triton X-100, dechorionated with 50% bleach, washed with NaCl/Triton and transferred to an ice-cold Dounce homogenizer. Embryos were then washed twice with cold H2O and once with cold lysis buffer. Embryos were suspended with 3 volumes of lysis buffer and lysed with 25 strokes of an ice-cold A pestle, followed by 25 strokes with a cold B pestle. Embryo lysate was transferred to microfuge tubes and centrifuged at 15,000 x g for 10′ at 4°. The supernatent was removed to fresh tubes and either used immediately or frozen in aliquots on dry ice and stored at -80.
For head lysates, flies of the appropriate genotype were anaesthetized with CO2, transferred to a Falcon tube and flash frozen with liquid N2. Immediately after the nitrogen boiled off, the flies were vortexed vigorously to shake off heads and appendages (taking care to keep the flies frozen), and heads were isolated away from bodies and legs with appropriate minisieves (Fisher): heads are passed by the #25 mesh but retained by the #45 mesh. Sieve assembly was prechilled with liquid N2 and kept cold during the separation process. Isolated heads were transferred to a puddle of liquid N2 in a prechilled agate mortar, ground to a fine powder with a chilled pestle, and transferred to an ice-cold Dounce. Ice-cold lysis buffer was added (using 1/5 the starting volume of whole flies, corresponding to ~3x the volume of isolated heads) and the powder was homogenized with 25 strokes each of the A pestle and B pestle, avoiding excessive frothing of the lysate. Lysate was clarified, frozen and stored as for embryo lysates.
Protein A sepharose beads (Amersham) were blocked with lysis buffer containing 0.5% BSA, prebound with Rb anti-mouse IgG (10μg Rb anti-mouse per 10 μl of beads), washed 2x, bound with the appropriate primary antibody, and washed 2x. Extract was thawed on ice, and 10μl of protein A /Rb anti-mouse beads was added per IP sample (150 μl of extract, diluted to 300μl final volume with lysis buffer) for pre-clearing. Fresh 0.2mM PMSF was added to thawed lysate prior to clearing. Sample was rocked at 4° for 1 hr, then cleared by centrifugation at 15,000 x g, 10′, 4°. Supernatent was removed and used immediately for co-immunoprecipitation. Beads bearing the appropriate primary antibody (10μl) were added per IP sample and rocked at 4° for 90′. Beads were spun down for 5″, unbound material was removed, and the beads were washed 4x with lysis buffer. Beads were then boiled in 20 μl Laemmli sample buffer. Samples were analyzed by SDS-PAGE and Western blotting. ECL detection was used, employing the Pierce SuperSignal West Pico or Femto reagents, or the Amersham ECL Advance system.
(A) Stage 16/17 wild type embryos, or (B) embryos deficient for disabled (Df(3L)std11), were collected, fixed, stained with polyclonal anti-Disabled antibodies and visualized by peroxidase immunocytochemistry. The pattern of tissues labelled by anti-Disabled in wild type matches the published pattern of Disabled immunoreactivity (Gertler et al., 1993); labelling above background is not observed in the mutant. Similar results were obtained with both monoclonals used in these studies (not shown).
(C) Extract of wild type embryos was subjected to immunoprecipitation with monoclonal (mAb P4D11G5; lane 2) or polyclonal (lane 3) anti-Disabled antibodies, separated by SDS-PAGE and analyzed by immunoblotting with anti-Disabled (MAb P6E11A7). Input extract is in lane 1. Bracket highlights the family of Disabled -immunoreactive species; circle indicates band due to precipitating antibody.
Immunoprecipitations from adult head extract were assayed by Western blotting with anti-Notch. Notch protein (arrow) is detected in anti-Trio immunoprecipitates of wild type extract (lane 3) and additional Notch protein is detected when anti-Trio precipitations are done from extracts of animals that overproduce Notch (elav-GAL4; UAS-N; lanes 4-6; compare lanes 3 and 4). A very small amount of Notch protein does precipitate non-specifically in control IP of extract from Notch-overproducing heads (lane 6), but it is far less than the co-IP with Trio (or Disabled) from the overproducing samples (lanes 4 and 5, respectively). Input extracts (lanes 1 and 2) are shown for comparison. Assaying the same input extracts by blotting with anti-β-tubulin (bottom panel) reveals that the total protein concentration in the wild type extract was at least as high (indeed slightly higher) than the total protein in the extract with overproduced Notch. Note: this is the same experiment shown in Fig 6C, but in a lighter ECL exposure to facilitate comparison of the immunoprecipitates from wild type and Notch-overproducing samples.
We wish to thank all the members of our lab for discussions and advice, and we particularly thank Lei Wang for help with co-IP experiments, and Dan Crowner and Lori Luke for outstanding technical assistance. For helpful discussions about the biochemical studies we thank Brian Howell, and for comments on the manuscript we thank Bob Callahan, Ken Irvine, Chi-Hon Lee, Ron McKay and Valerie Vasioukhin. We also thank the many members of the fly community who generously provided published and unpublished reagents and shared information and advice, particularly Chihiro Hama, Alfonso Martinez-Arias, Marc Muskavitch and Cedric Wesley. Essential antibodies were obtained from the Developmental Studies Hybridoma Bank, and flies from the Bloomington Drosophila Stock Center. Liz Wayner and the staff of the FHCRC Antibody Development Facility, as well as Doug Woodle and the FHCRC Biologics Production Facility are gratefully acknowledged for assistance with antibody development and production. Early portions of this work were performed while the authors were at the Fred Hutchinson Cancer Research Center, with support from NIH grant GM57830, and this research was funded, in part, by the Intramural Research Program of the NIH, NINDS.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
We raised both polyclonal and monoclonal mouse antibodies against Drosophila Disabled, and characterized their activity and specificity. Anti-Disabled antibodies labelled a pattern of tissues in wild type embryos that reproduced the published tissue distribution of Disabled (Fig S-1A; compare (Gertler et al., 1993)), but did not significantly label homozygous deficiency embryos (Fig S-1B). Moreover, in Western blots of wild type embryo extract the anti-Disabled antibodies recognized a family of high molecular weight bands resembling the published pattern of Disabled protein products (Fig 5A, bottom panel; Fig S-1C). These are thought to represent both multiple splice forms and phosphorylation states of Disabled (Gertler et al., 1993). The same family of bands was immunoprecipitated by both the polyclonal and monoclonal antibodies (Fig S-1C), and none of these bands was detected in extract of deficiency embryos (data not shown).