The Adam family metalloprotease Kuzbanian regulates the cleavage of the roundabout receptor to control axon repulsion at the midline

Slit ligands and their Robo receptors play conserved roles in regulating repulsive axon guidance during nervous system development (Dickson and Gilestro, 2006; Garbe and Bashaw, 2004; Nguyen-Ba-Charvet and Chedotal, 2002). At the midline in Drosophila, Slit is secreted by midline glia and, through the activation of Robo receptors, prevents abnormal crossing of the midline by ipsilateral axons and re-crossing by commissural axons (Kidd et al., 1999; Kidd et al., 1998a). Loss-of-function mutations in either slit or robo lead to axon misrouting at the midline: slit mutations result in the complete collapse of all CNS axons, whereas mutations in robo result in a milder phenotype in which axons cross and re-cross the midline many times (Battye et al., 1999; Kidd et al., 1999; Kidd et al., 1998a).

Slit proteins are large secreted proteins consisting of leucine-rich repeats (LRRs), seven to nine EGF repeats and a C-terminal cysteine knot (Brose and Tessier-Lavigne, 2000; Rothberg et al., 1990). In both fly and vertebrate systems, Slit proteins undergo proteolytic processing to generate a large N-terminal fragment (Slit-N) and a smaller C-terminal fragment (Brose et al., 1999; Wang et al., 1999). Experiments using cultured rat dorsal root ganglion or olfactory bulb neurons indicate that different fragments of Slit have different properties (Nguyen-Ba-Charvet et al., 2001). Full-length Slit (Slit-FL) and Slit-N bind Robos and repel axons, but the two forms have opposite effects on axon branching: Slit-N stimulates branching and Slit-FL inhibits branching (Nguyen-Ba-Charvet et al., 2001).

What molecules might contribute to Slit processing? Kuzbanian (Kuz) was originally identified in Drosophila for its role in regulating Notch signaling during neurogenesis (Pan and Rubin, 1997; Rooke et al., 1996). Kuz is a single-pass transmembrane metalloprotease belonging to the Adam family that is widely expressed throughout development in the Drosophila central nervous system (Fambrough et al., 1996). It is expressed at the cell surface where it recognizes and cleaves its substrates within their extracellular domain, resulting in ecto-domain shedding. Although several Kuz substrates have been identified – including Notch, APP (amyloid precursor protein; also known as Appl – FlyBase) and ephrins – protein sequence analysis has not revealed a signature cleavage or Kuz association site, making it difficult to predict Kuz substrates by sequence alone (Becherer and Blobel, 2003; Gomis-Ruth, 2003). Kuz processes substrates whose functions range from cell fate specification to axon guidance and these processing events are essential for the correct development of the CNS (Yang et al., 2006).

In the context of axon guidance, the mammalian Kuz homolog ADAM10 has been shown to regulate the cleavage of Ephrin A2 ligands during Ephrin-dependent contact repulsion (Hattori et al., 2000). ADAM10 is constitutively associated with the GPI-linked Ephrin A2 ligand. Upon binding of the ligand by the EphA2 receptor, Kuz cleaves Ephrin A2 in the extracellular juxta-membrane region, releasing it from the membrane and allowing the growth cone of the EphA2 expressing cell to retract (Hattori et al., 2000). In addition, expression of a dominant-negative form of Kuz in Drosophila midline glia results in ectopic crossing of ipsilateral axons, and there are dose-dependent genetic interactions between kuz and slit, suggesting that Kuz might regulate the cleavage of Slit during midline axon repulsion (Schimmelpfeng et al., 2001).

Here, we report that, in a sensitized screen for genes involved in Slit-Robo repulsion, we have identified several alleles of kuz. We find no evidence that Kuz is involved in the processing of Slit, nor does Slit proteolysis appear to be required for the normal repulsive guidance function of Slit. Genetic rescue experiments demonstrate that although Kuz is normally expressed in both midline glia and CNS neurons, neuronal expression of Kuz completely rescues kuz mutant phenotypes, whereas midline expression does not. We also present evidence that Kuz promotes cleavage of the Robo receptor in vitro and that an uncleavable form of the Robo receptor (Robo-U) is unable to rescue the robo mutant phenotype, suggesting that it does not maintain normal Robo repulsive activity. Finally, we show that ADAM10/Kuz function is required for the Slit-dependent recruitment of the Sos (Son of Sevenless) Ras/Rac GEF to the Robo receptor, suggesting that Kuz is important for Robo receptor activation. Together, these observations support the hypothesis that Kuz-directed cleavage of Robo is important for axon repulsion at the midline.

The following mutant alleles were used in this study: comm^Δe39, robo^GA285, robo^z1772, slit², kuz^e29, kuz^H143, kuz¹¹², kuz⁴²⁰ and kuz²⁵⁸³. All robo, slit and kuz alleles were balanced over CyWgβGal. To generate transgenic fly strains, UASKuzHA, UASSlitU and UASRoboU were transformed into w¹¹¹⁸ flies using standard procedures. The Gal4-UAS system was used to express transgenes in the apterous ipsilateral neurons (apGal4), in the eagle commissural neurons (egGal4), in all neurons (elavGal4), or in midline glia (slitGal4 and simGal4). All crosses were conducted at 25°C.

To generate uncleavable Slit, 27 base pairs encoding HNMISMMYP between the fifth and sixth EGF repeats were deleted using standard procedures. All other constructs were made using standard molecular biology techniques.

S2R+ cells on poly-L-lysine coated plates were transfected at 40% confluency using Effectene (Qiagen) and induced 24 hours later with 0.5 mM copper sulfate. Twenty-four hours after induction, the culture media was harvested and the cells were lysed in 1 × PBS, 0.5% Triton X-100, 1 × protease inhibitor (Roche) for 20 minutes at 4°C. Proteins were resolved on SDS-PAGE gels and blotted with rabbit anti-GFP (Molecular Probes, 1:500), mouse anti-Robo (DSHB, 1:50), mouse anti-HA (BabCO, 16B12, 1:1000) or mouse anti-Myc (9E10, 1:250). The following secondary antibodies were used: rabbit anti-HRP and mouse anti-HRP (Jackson Laboratories). Blots were developed with the ECL Plus Western Blotting Detection System (Amersham).

dsRNA was made using standard techniques. PCR amplification of Drosophila kuz was conducted using primers that included 5 ′ T7 promoter regions: 5 ′-TTAATACGACTCACTATAGGGAGACACCAGTACTCCGTT-3 ′ (forward) and 5 ′-TTAATACGACTCACTATAGGGAGTTCGCTATGTGGCGCCA-3 ′ (reverse). Reverse transcription was performed using the MEGAscript T7 High Yield Transcription Kit (Ambion). Single-stranded sense and anti-sense RNAs were incubated in 100 mM potassium acetate, 30 mM HEPES-KOH (pH 7.4), 2 mM magnesium acetate for 4 minutes at 95°C, followed by 10 minutes at 70°C, then slowly cooled to 4°C. S2R+ cells were transfected and processed as described above and 5 μl of the double-stranded RNA was added at the time of transfection.

HRP immunohistochemistry was performed as previously described, and images were obtained using a Zeiss Axiocam and Openlab software (Improvision). For fluorescence staining, the following antibodies were used: mouse 1D4/FasII [Developmental Studies Hybridoma Bank (DSHB), 1:100], mouse MAb BP102 (DSHB, 1:100), rabbit anti-Myc (Sigma-Aldrich, 1:500), mouse anti-Slit (DSHB, C555.6D, 1:25), mouse anti-βgal (DSHB, 40-1a, 1:250), rabbit anti-GFP (Molecular Probes, 1:500), mouse anti-Robo (DSHB, 1:50), mouse anti-HA (BabCO, 16B12, 1:1000), and mouse anti-Myc (9E10, 1:250). Stacks of images were obtained using a Leica DMIRE2 confocal and a 63 × oil immersion objective. A maximum projection of the stacks was generated with NIH Image/ ImageJ software.

293T cells were seeded on poly-Lysine coated coverslips and transfected at 40% confluency using Effectene (Qiagen). Twenty-four hours after transfection, cells were starved in serum-free DMEM medium for 12-16 hours and then stimulated with conditioned medium of hSlit2-stably-expressing 293T cells (a gift from Dr Y. Rao, Peking University School of Life Sciences, Beijing, China) for 5 minutes. Treated cells were washed with 1 ×PBS once and immediately fixed in 4% paraformaldheyde/1 ×PBS for 20 minutes. Fixed cells were permeabilized in 0.1% Triton X-100/1 ×PBS for 2 minutes and blocked with 3% BSA/1 ×PBS for 5 minutes. Cells were then incubated with primary antibody (rabbit anti-hSos, 1:50; mouse anti-Myc, 1:1000; rat anti-HA, 1:1000) overnight at 4°C and then secondary antibody (rabbit AlexaFluor488, mouse Cy3, rat Cy5 secondary antibody) for 30 minutes at room temperature. For each group of cells, at least 10 cells were randomly selected for quantification. Average fluorescence intensity was calculated using NIH Image J software, as previously described (Yang and Bashaw, 2006).

In wild-type Drosophila embryos, staining with the BP102 monoclonal antibody (MAb) revealed a ladder-like structure of axons composed of longitudinal connectives that are bridged by anterior and posterior commissures in each segment (Fig. 1A). The majority of axons cross the midline once and only once, while a smaller set of ipsilateral axons stay on their own side of the midline (Dickson and Gilestro, 2006). Both kuz and robo mutants showed thinning of the longitudinal connectives paired with thickening of the commissures (Fig. 1B,C). In robo mutants, this phenotype results from loss of midline repulsion, which allows axons to abnormally re-enter the midline (Kidd et al., 1998a; Seeger et al., 1993). The similarity of phenotypes observed in kuz and robo mutants suggests that Kuz might be involved in midline repulsion.

In wild-type embryos, Fasciclin II (FasII)-positive axons project longitudinally and never cross the midline (Fig. 1E), whereas mutations in either robo or kuz result in differing degrees of abnormal midline crossing (Fig. 1F,G). We took advantage of the mild midline crossing defects associated with simultaneous 50% reduction of the slit and robo genes (Fig. 1I) to perform a genetic screen to identify additional components of the repulsive guidance pathway. We screened a collection of several hundred mutants that were previously isolated in a large screen for genes that regulate the formation of axon commissures (Seeger et al., 1993). Several alleles of kuz were found to dominantly enhance the slit, robo transheterozygous phenotype (Fig. 1J; see Fig. S1 and Table S1 in the supplementary material), suggesting that kuz is involved in midline repulsion and might be a positive regulator of slit-robo signaling. In addition, removing all of the zygotic kuz in slit, robo trans-heterozygotes enhanced the kuz mutant phenotype, whereas kuz, robo double mutants exhibited a midline crossing phenotype that was no stronger than that of robo mutants alone (Fig. 1D,H; see also Table S1 in the supplementary material; data not shown). Here, it is important to note that kuz zygotic mutants maintain significant levels of kuz function owing to maternally deposited genes/proteins and that they do not exhibit cell fate defects (see Fig. S2 in the supplementary material) (Fambrough et al., 1996); the loss of both maternal and zygotic kuz leads to notch mutant cell fate specification phenotypes (Rooke et al., 1996).

To obtain further evidence for a role of kuz in Slit-mediated repulsion, we analyzed embryos that were mutant for kuz and commissureless (comm). Comm normally functions to inhibit Robo repulsion by preventing delivery of the Robo receptor to the growth cone plasma membrane (Keleman et al., 2002; Keleman et al., 2005), and mutations in comm result in a complete absence of axon commissures because of excess Robo repulsion (Fig. 1K) (Kidd et al., 1998b). As comm, robo double mutants resemble robo single mutants (Seeger et al., 1993), we reasoned that, if kuz promotes Robo repulsion, then at least some axons should be able to cross the midline in kuz; comm double mutants. This is indeed the case; significant restoration of commissure formation was observed in double mutants, although it was apparent that most axons still failed to cross the midline (Fig. 1L). Here, we note that this degree of suppression probably underestimates the role of kuz, as there is a significant maternal contribution. Together, these genetic results support the idea that Kuz functions in the Slit-Robo pathway to regulate midline repulsion.

We next sought to more closely characterize the behavior of a smaller group of ipsilateral neurons, the Apterous (ap) neurons (Lundgren et al., 1995). Removing either one copy of kuz or one copy of slit resulted in a mild ectopic crossing defect, whereas simultaneously limiting kuz and slit significantly enhanced this phenotype (see Fig. S1C in the supplementary material). Additionally, restricted misexpression of a dominant-negative kuz transgene (Pan and Rubin, 1997) in ap neurons caused ectopic crossing of axons in about fifteen percent of segments (Fig. 2B,D), and this phenotype was significantly more severe in slit or robo heterozygotes (Fig. 2C,D; see also Table S1 in the supplementary material). These data solidify the idea that slit and kuz interact genetically, and also suggest that kuz function is required in neurons.

Previous reports have shown that kuz mRNA is broadly expressed throughout the developing embryonic CNS, including in neurons and midline glia (Fambrough et al., 1996). Because we observed a discrepancy between previous studies suggesting that Kuz protease activity is required in midline glia (Schimmelpfeng et al., 2001) and our finding that Kuz function is required in the ap neurons, we sought to further elucidate in which cells kuz expression is necessary. To test where kuz is required for midline guidance, we performed genetic rescue experiments using the Gal4-UAS system (Brand and Perrimon, 1993). Expression of UASKuz or tagged UASKuzHA in all postmitotic neurons using elavGal4 completely rescued the ectopic midline crossing of FasII-positive axons in kuz mutants (Fig. 3A,B,D,F; data not shown) (Fambrough et al., 1996). By contrast, expression of UASKuz or UASKuzHA in midline glia using slitGal4 did not rescue ectopic midline crossing (Fig. 3C,E). Although it is difficult to directly compare levels of expression in the two different cell types, antibody staining to detect transgenic UASKuzHA revealed that the two drivers appear to express comparable levels of Kuz (Fig. 3E,F). Taken together with the observation that expression of dominant-negative Kuz in the ap ipsilateral neurons causes ap axons to ectopically cross the midline (Fig. 2B), these rescue data further support the idea that Kuz is required in neurons for repulsion. Although these observations point to a potential cell autonomous role for kuz in regulating Robo repulsion, it is also possible that Kuz could act non-autonomously by cleaving substrates in trans on adjacent cells. Indeed, ADAM10 has been shown to act non-autonomously to cleave Ephrin A5, in trans, on adjacent cell surfaces (Janes et al., 2005).

In order to directly test the importance of Slit processing for midline repulsion, we deleted the nine amino acids that make up the spacer region between the fifth and sixth EGF repeats to generate an uncleavable version of the protein-Slit-U (Fig. 4A). Biochemical analysis using antibodies directed against either the N- or C-terminus of Drosophila Slit revealed that indeed Slit proteolysis in cultured 293T cells and in Drosophila embryos is abolished by this deletion (Fig. 4B,C). We next compared the function of Slit-U with that of wild-type Slit (Slit-FL) in a transgenic rescue assay. In slit mutant embryos, all axons collapsed on the midline (Fig. 4D,F). Restoration of Slit-FL expression in the midline glia using slitGal4 provided a strong rescue of the ectopic crossing phenotype seen in slit mutant embryos, but did not appear to restore proper lateral position of the FasII-positive fascicles (Fig. 4D,G). Surprisingly, we found that midline expression of Slit-U was able to rescue the midline guidance phenotypes of slit mutants just as well as expression of wild-type Slit-FL (Fig. 4D,H). Several independent transgenic lines gave similar results in our rescue assay (data not shown). The fact that the effects of Slit-U in rescue experiments were indistinguishable from those of Slit-FL suggests that the cleavage of Slit is not essential for its role in midline repulsion.

Because, in the case of Notch signaling, Kuz has been shown to promote the cleavage of both the Delta ligand and the Notch receptor (Pan and Rubin, 1997; Qi et al., 1999; Six et al., 2003), we next investigated the possibility that the Robo receptor might be a substrate for Kuz. Given that Kuz is a transmembrane protein with extracellular protease activity, we first tested whether we could observe the release of the Robo extracellular domain (Robo-ECTO) into the media from Robo-transfected cells and found that we could readily detect low levels of an approximately 100-kDa Robo immunoreactive fragment in the culture media (Fig. 5A, lane 1). Importantly, this fragment was not observed in untransfected cells (Fig. 5A, lane 6). To further verify that this fragment corresponds to Robo-ECTO, we generated HA-tagged Robo and assayed the culture media by immunoblotting with anti-HA. Again, we detected a 100-kDa fragment, indicating that indeed this represents the Robo-ECTO that is shed from the cell surface (Fig. 5B, lane 1).

We next tested whether co-expression of Kuz could influence the production of this cleavage fragment. Indeed, in cells co-transfected with Kuz and Robo, there was an increased amount of Robo-ECTO detected in the media (Fig. 5A, lanes 4 and 5), suggesting that Robo could be a substrate of Kuz in vitro. Expression of a form of Kuz that lacks the protease domain (KuzΔmetalloprotease) had no influence on the accumulation of Robo-ECTO (Fig. 5A, lanes 2 and 3). Antibody staining of these cells revealed comparable levels of membrane-localized Kuz or KuzΔmetalloprotease, as well as similar degrees of colocalization with Robo (see Fig. S3 in the supplementary material). These data demonstrate that Robo-ECTO shedding can be enhanced by Kuz in vitro.

To further demonstrate the Kuz dependency of Robo-ECTO shedding, we performed RNA interference experiments to show that knockdown of Kuz prevents the enhanced shedding (Fig. 5C). Treatment with kuz double-stranded RNA (dsRNA) significantly reduced the levels of KuzHA when compared with those of untreated cells (Fig. 5C, lanes 3-6). Comparison of Robo-ECTO levels in the media of cells treated with kuz dsRNA (lanes 1 and 3) with levels in untreated cells (lanes 2 and 4) showed that attenuation of kuz expression results in a marked decrease in Robo-ECTO in the media, both in the presence and in the absence of transfected Kuz (Fig. 5C). We were unable to unambiguously demonstrate that kuz dsRNA treatment knocks down expression levels of endogenous Kuz because there is no antibody specific to Drosophila Kuz. However, the observation that kuz dsRNA treatment results in a decrease in Robo-ECTO shedding is highly suggestive that this treatment is able to knock down endogenous Kuz levels (Fig. 5C, lanes 1 and 2).

We next sought in vivo evidence for a role of kuz in regulating Robo processing. Specifically, we examined the surface expression levels and localization of Robo in the embryonic CNS of live-dissected kuz mutant and sibling embryos (Fig. 6). In contrast to kuz/+ heterozygotes, in which surface Robo expression was restricted to the longitudinal portions of CNS axons (Fig. 6A-C), kuz mutants showed elevated expression of Robo and a marked mislocalization of Robo on the surface of axon commissures (Fig. 6D-F). These observations are consistent with a previous report showing mislocalization of Robo in fixed and permeabilized kuz loss-of-function embryos (Schimmelpfeng et al., 2001). Quantification of Robo surface levels revealed a significant increase in expression that is consistent with an increase in the detection of full-length Robo by western blot (Fig. 6G,H). Despite the increase in Robo expression levels, we were unable to consistently detect the presence of ectodomain fragments in wild-type embryos, potentially because of their rapid degradation or because of our assays were not sensitive enough. This limitation precluded the analysis of the effects of kuz mutations on the in vivo cleavage of Robo. Nevertheless, together with our genetic interaction data and in vitro biochemical data, these observations further suggest that Kuz-dependent cleavage of Robo is important in vivo for midline repulsion.

Both our genetic and biochemical evidence suggest that Kuz processing of Robo is important for Robo repulsive function in the context of midline guidance. In order to directly test this hypothesis in vivo, we created an uncleavable form of the Robo receptor (Robo-U). Our initial efforts to create an uncleavable Robo via small insertions were not successful; however, we were able to create an uncleavable Robo receptor by swapping the first three Fibronectin type III (FnIII) domains from the attractive Netrin receptor Frazzled (Fra) for the three Fn domains of Robo (Fig. 7A). It is important to note that the region of Fra used to replace the three Fn domains of Robo does not include the Netrin binding domain, therefore the activity of Robo should not be affected by the presence of Netrin in vivo. We chose Fra as a swapping partner to create Robo-U because it has a similar structural organization to Robo. In addition, we did not detect Kuz-dependent shedding of the Fra ECTO domain in Fra-transfected cells with an antibody raised against the N-terminus of Fra (Fig. 7B, lanes 4 and 8). We were unable to detect shed ectodomain in the media of cells expressing Robo-U, either with or without added Kuz, suggesting Robo-U is not cleaved in this assay (Fig. 7B, lanes 3 and 7). We were also interested in determining whether Fra-Ro (a chimeric receptor consisting of the ectodomain of Fra and the transmembrane and intracellular domains of Robo) would undergo ectodomain shedding, because previous studies have shown that it is capable of mediating Netrin-dependent repulsive signaling in in vivo gain-of-function assays (Bashaw and Goodman, 1999). Interestingly, Fra-Ro does not seem to undergo ectodomain shedding in vitro, suggesting that its in vivo repellent activity might be independent of Kuz function (Fig. 7B, lanes 2 and 6). All four of the receptors assayed in the ectodomain shedding experiment were properly expressed at the plasma membrane of Drosophila S2 cells, suggesting that they are normally trafficked (see Fig. S3 in the supplementary material). In addition, we confirmed that Robo-U retains Slit binding activity that is qualitatively similar to that of wild-type Robo using a cell overlay binding assay (see Fig. S3 in the supplementary material).

Having determined that Robo-U is not cleaved in our in vitro assay, we next sought to characterize the activity of this receptor in vivo. Loss of Robo function resulted in a severe phenotype in which the FasII-positive ipsilateral axons repeatedly crossed the midline (Fig. 7E). This phenotype could be almost completely rescued by expressing UASRobo with the pan-neural driver elavGal4 (Fig. 7H) (Fan et al., 2003; Garbe and Bashaw, 2007). We reasoned that if Robo-U maintained its signaling function in vivo, expressing it in robo mutant embryos would rescue the ipsilateral crossing phenotype, as wild-type Robo does. However, pan-neural expression of UASRobo-U in robo mutant embryos provided only modest rescue of the mutant phenotype, with many ipsilateral axons still crossing the midline (Fig. 7F), which suggests that Robo processing is important for its repulsive signaling function at the midline. Several independent inserts of the Robo-U transgene gave similar results (Fig. 7C). Fra-Ro also provided only a very slight rescue of the robo phenotype (Fig. 7G), suggesting that despite its ability to mediate repulsive activity when misexpressed in a gain-of-function assay (Bashaw and Goodman, 1999), it cannot substitute for endogenous robo in midline repulsion. Importantly, transgenic expression levels of Robo-U, Fra-Ro and Robo were comparable, as assessed by immunostaining for the Myc epitope tag (Fig. 7F-H, bottom panels; see also Fig. S4 in the supplementary material). Interestingly, although transgenic wild-type Robo was cleared from commissures as endogenous Robo was, Robo-U was expressed along the entire length of the commissural axons (Fig. 7F-H, arrowheads; see also Fig. S4 in the supplementary material). It is possible that proteolysis of Robo is important for its exclusion from the commissural portion of contralateral axons. This observation is also consistent with our surface staining experiments (Fig. 6), as well as with a previous report showing that crossing axons in kuz loss-of-function embryos express Robo protein (Schimmelpfeng et al., 2001).

In order to quantify the level of rescue, we counted the number of segments in which the ipsilateral ap axons ectopically crossed the midline. In robo embryos the ap axons completely collapsed on the midline (Fig. 7J). Transgenic expression of either Robo-U or Fra-Ro in ap neurons showed comparably low levels of rescue with respect to ectopic midline crossing when compared with the near complete rescue seen in robo embryos expressing Robo (Fig. 7C,K-M). However, in contrast to the complete collapse of ap axons in robo embryos, expression of these chimeric receptors did seem to partially restore midline repulsion, with some segments having ap axon tracts on both sides of the midline (Fig. 7K-M). The three transgenic receptors assayed each showed comparable levels of expression in the ap neurons (Fig. 7K-M, bottom panels).

Although the inability of Robo-U to restore proper repulsion at the midline supports the hypothesis that processing of Robo is important for this process, it is also possible that this chimeric receptor has simply lost its signaling ability as a result of misfolding or some other alteration of function due to exchanging the Fn repeats. To test whether Robo-U is able to mediate repulsion in vivo, we misexpressed this receptor in a small subset of contralateral neurons, the eagle (eg) neurons. In wild-type embryos, the eg neurons crossed the midline in both the anterior and posterior commissures of each segment (see Fig. S5A in the supplementary material). Misexpression of UASRobo in eg neurons prevented the posterior subset of eg axons from crossing the midline because of the high levels of repulsion mediated by excess Robo signaling (see Fig. S5D in the supplementary material). UASFra-Ro misexpression showed a similar phenotype. Misexpression of UASRobo-U was also able to repel the posterior eg neurons from the midline (see Fig. S5B in the supplementary material), suggesting that it maintains repulsive signaling activity in vivo. This repulsive activity is seemingly independent of Robo processing and is likely to be initiated by a different mechanism to that used to activate endogenous Robo. It is possible that over or misexpression of these receptors causes them to behave in a way that they would not normally in vivo. We believe that our rescue assay is a better determinant of how these transgenic receptors act in the context of normal development, as we are directly measuring their ability to substitute for the endogenous receptor. These rescue experiments suggest that cleavage of Robo is important for its repellant activity in the context of midline guidance; however, we cannot exclude the possibility that Robo-U could be deficient for other non-cleavage dependent functions.

If Robo processing by Kuz is important for Robo activation, disruption of Kuz activity might prevent the association of downstream signaling molecules with the cytoplasmic domain of Robo. Sos is a Ras/Rac guanine nucleotide exchange factor (GEF) that associates with the Slit-bound Robo receptor in a ternary complex with the SH3-SH2 adaptor protein Dreadlocks to regulate Rac activity (Yang and Bashaw, 2006). Studies in mammalian cells show that the Rac-GEF activity of Sos is highly dependent upon its precise subcellular localization (Innocenti et al., 2002), suggesting that if Sos recruitment to the plasma membrane was blocked, its activity would be disrupted. In order to determine whether Kuz regulates the Slit-dependent recruitment of Sos to the plasma membrane, we used the mammalian HEK293T system. In the absence of the ligand Slit, endogenous Sos was located diffusely in the cytosol in cells transfected with human ROBO1 (hROBO1), and the cultured cells had a splayed-out flattened morphology (Fig. 8A-D). Bath application of human SLIT2 (hSLIT2) to hROBO1-transfected cells induced a recruitment of Sos protein to the membrane and a distinct rounded morphology (Fig. 8E-H). By contrast, expression of a dominant-negative form of mammalian ADAM10 was able to block the Slit-induced relocalization of Sos and its associated change in cell morphology (Fig. 8I-L; see also Fig. S6 in the supplementary material), suggesting that Kuz processing of the Robo receptor is an important step in the initiation of its signaling cascade. Importantly, ADAM10DN was appropriately expressed at the plasma membrane (see Fig. S6 in the supplementary material). Expression of ADAM10DN in the absence of hSLIT2 treatment did not affect the gross morphology of the cells, demonstrating that expression of this protein does not interfere with overall cell health (see Fig. S6 in the supplementary material). These data provide evidence that limiting Kuz/ADAM10 activity in Robo-expressing cells results in a significant reduction in the ability of Robo to initiate downstream signaling events in response to Slit stimulation and further suggest that Kuz/ADAM10 regulation of Robo signaling is conserved in humans.

Both our genetic and biochemical findings support the hypothesis that cleavage of the Robo receptor – rather than its Slit ligand – by the metalloprotease Kuz is important in the context of midline guidance. Loss of Kuz protease activity or of the cleavage site of Robo in vivo results in ectopic crossing of ipsilateral axons because of the loss of Robo-mediated repulsion, whereas an uncleavable form of Slit is able to rescue guidance defects in slit mutants as well as does Slit-FL. Furthermore, biochemical analyses have demonstrated that Robo is a substrate of KuzADAM10 in vitro. Finally, our Sos recruitment assay demonstrates that reduction of endogenous Kuz protease activity attenuates Slit-dependent relocalization of Sos to the plasma membrane, where it acts as a regulator of actin cytoskeletal rearrangement and, presumably, growth cone retraction.

Our data suggest a model in which Kuz promotes Robo ectodomain shedding as a mechanism of Robo activation (Fig. 9). We propose that Kuz cleavage of Robo is initiated by binding of Slit, and that the release of the ectodomain of Robo causes a conformational change in Robo that allows its cytoplasmic domain to associate with Sos via the SH3-SH2 adaptor protein Dreadlocks. Sos is then properly localized in order to exert its effect on cytoskeletal rearrangement.

In light of the strong evidence from vertebrate studies indicating that the different Slit cleavage products have distinct properties (Nguyen-Ba-Charvet et al., 2001), we were surprised to find that an uncleavable form of Slit can rescue slit mutants as effectively as can wild-type Slit. What then is the significance of Slit cleavage? Although the cleavage fragments are clearly present in western blots of total embryonic protein, it is not known where in the embryo this cleavage is occurring. Proteolysis might be important for developmental events other than axon guidance that involve Slit; for instance, muscle migration or attachment (Kidd et al., 1999; Kramer et al., 2001), or formation of the heart (Qian et al., 2005; Santiago-Martinez et al., 2006). Future experiments might shed light on the significance, if any, for Slit proteolysis in these contexts. Even though Slit-FL and Slit-U appear to be largely interchangeable in the experiments presented here, it cannot be ruled out that Slit proteolysis plays a role in fine-tuning axon guidance. Slit-FL does not fully rescue slit mutant axon guidance defects, which leaves open the question of whether cleavage is important for those guidance events that are not rescued; for example, the fine-tuning of the lateral positioning of axons.

Although it seems evident that Kuz activity is important for Robo-mediated growth cone retraction, it is unclear how Robo ectodomain shedding is involved in the repulsive process. Both Notch and ephrins are known to be substrates of Kuz, but the role that Kuz plays in their signaling is very different. GPI-linked Ephrin A2 forms a stable complex with ADAM10, although ADAM10 proteolytic activity is only initiated when EphA3, the transmembrane Eph receptor, is present. ADAM10 cleavage of Ephrin A2 can be considered a permissive event, in that it releases the strong Eph-ephrin tether that attaches the two cell surfaces, thereby allowing the EphA3-expressing growth cone to retract (Hattori et al., 2000). The role of Kuz in Notch signaling is more directly linked to Notch activation (Mumm and Kopan, 2000). Although the genetic data we present cannot distinguish whether Kuz acts in a permissive or an activating capacity with respect to Robo signaling, the observation that the expression of dominant-negative ADAM10 blocks Slit-induced recruitment of Sos to the plasma membrane suggests that Kuz/ADAM10 is likely to be important for the association of Robo with its signaling effectors. In other words, it appears that Kuz/ADAM10 contributes to the initiation of Robo signaling events.

If Kuz is indeed playing an activating role in Robo signaling, it should be regulated in a way to prevent continuous repulsive signaling. The most parsimonious explanation for regulation of Kuz activity is that it is Slit dependent. Indeed Notch and Ephrin proteolysis by Kuz is known to be dependent upon ligand binding. Additionally, other studies have demonstrated that ADAM10 substrates, including APP and Notch, are cleaved upon receptor-ligand binding (Beel and Sanders, 2008). We tested whether Kuz proteolysis of Robo was also dependent upon ligand binding, but unfortunately we were not able to detect a Slit-induced effect on Kuz-dependent Robo ectodomain shedding in vitro (data not shown). However, these experiments were performed in Drosophila S2 cells in which both Robo and Kuz were overexpressed, and the normal regulation of cleavage might not be maintained in this context. The possibility also exists that Kuz processing of Robo might be regulated by calcium influx, differential substrate glycosylation events, or substrate oligomerization, as is observed with some ADAM10 substrates (Beel and Sanders, 2008). In the future, it will be important to determine if Robo proteolysis is dependent on Slit binding, perhaps by examining, both in mammalian cells and in vivo, the processing of a Robo receptor that cannot bind Slit.

We thank Kim Bland and Corey Goodman for generating the uncleavable Slit construct, John Flanagan for the mouse ADAM10 reagents, D. J. Pan for the Kuzbanian dominant-negative transgenes and the Bloomington Stock Center for Drosophila strains. We are grateful to members of the Bashaw lab for thoughtful discussion and to Andy McClelland for assistance with the quantification of the Sos recruitment assay. This work was supported by a Whitehall Foundation Research Grant, a Burroughs Wellcome Career Award, and NIH grants NS046333 and NS054739 to G.J.B. Deposited in PMC for release after 12 months.