A non-signaling role of Robo2 in tendons is essential for Slit processing and muscle patterning

In both vertebrates and invertebrates, the establishment of a functional musculoskeletal tissue requires a precise encounter between muscles and tendons (Schejter and Baylies, 2010; Schnorrer and Dickson, 2004; Schweitzer et al., 2010; Volk, 1999). Active cross-talk between tendons and elongating muscles has been demonstrated in Drosophila, where an important tendon-specific guidance signal is the secreted protein Slit (Kramer et al., 2001; Ordan et al., 2015; Wayburn and Volk, 2009). The contribution of the different Slit receptors, namely Robo1 (also known as Robo), Robo2 and Robo3, to muscle patterning remains elusive. Genetic studies have suggested that Robo1 and Robo2 cooperate in responding to the secreted tendon-specific ligand Slit on the muscle side (Kramer et al., 2001); however, this has never been directly tested.

Slit contains a conserved cleavage site (Brose et al., 1999; Wang et al., 1999). Recently, we showed that, in tendons, full-length Slit (Slit-FL) undergoes cleavage, creating an active, stable N-terminal polypeptide (Slit-N), which associates with the tendon cell surface, thus limiting the range of Slit signaling. Moreover, Slit cleavage at the surface of tendon cells was found to be crucial for short-range muscle repulsion, as well as for providing a stop signal for LT and DA3 muscles (Ordan et al., 2015); however, the molecular mechanism promoting localized Slit cleavage has yet to be elucidated. Here, we show that, for muscle guidance, Robo2 activity is required in tendon cells and not in muscles. Furthermore, Robo2 activity in tendons can promote Slit cleavage, creating the non-diffusible Slit-N; for this function the receptor extracellular domain is sufficient. Slit-N polypeptides were found to oligomerize, and this explains their reactivity with both tendons and muscles. Consistent with the role of Robo2 in muscle directional elongation by limiting Slit diffusion, knocked-in membrane-bound, full-length, uncleavable Slit rescued the muscle phenotype of robo2 mutants, whereas Slit lacking the membrane domain did not. These findings demonstrate a novel signaling-independent role for Robo2 in facilitating Slit activity in cis, which is important for promoting Slit short-range signaling.

To determine the individual contribution of Robo family receptors to the embryonic muscle pattern, we analyzed the orientation of the three lateral transverse muscles (LT1-3) and of the dorsal acute muscle 3 (DA3) in Drosophila embryos at stage 16 lacking distinct Robo receptors and compared it with that of slit mutants. In slit mutant embryos, the LT muscles are closer to the posterior segment border and are not spaced correctly, and the DA3 muscle does not extend its leading edge in an orthogonal fashion (Fig. 1A-B′) (Ordan et al., 2015).

The orientation of the LT and the DA3 muscles in robo1 or robo3 mutants exhibited almost normal positioning relative to their wild-type (WT) counterparts (Fig. 1C,E,C′,E′). Quantification of the distance of LT3 muscle to the posterior segment border divided by overall segmental width (D_LT3/D_s) revealed no significant difference between robo1 and WT (P=0.8313) or robo3 and WT (P=0.74) (Fig. 2L). However, in robo1;robo3 double-mutant embryos the orientation of these muscles resembled that of slit mutants (Fig. 1F,F′ and Fig. 2L; P=0.0001). robo2 mutants exhibited a phenotype that was comparable to that of slit mutants; the LT muscles were not equally spaced and were closer to the posterior segmental border, a phenotype observed in 85% of the segments (P=0.0001, n=47) (Fig. 1D). Quantification of D_LT3/D_s revealed a smaller value in the robo2 mutant (0.16 as compared with 0.34 in WT, P=0.0001; Fig. 2L). In addition, in 33% of the robo2 mutant segments, the DA3 muscle lost its diagonal orientation and often migrated in a straight ventral-dorsal orientation (P=0.0037, n=43; Fig. 1D′).

To reveal the dynamic nature of the process of muscle elongation in robo2 mutants, we followed individual muscles in live robo2 mutant and control embryos using specific GFP-marked LT (Fig. 1G,G′, Movies 1 and 2) or DA3 (Fig. 1H,H′, Movies 3 and 4) muscles at stages 13-16 throughout muscle elongation. This analysis showed that the polarity of the mutant muscles was similar to that of WT control muscles; however, the LT muscles tended to elongate closer to the segment border and were not evenly spaced. The DA3 muscle often did not turn towards the posterior segment, and both muscles reached their target tendons more slowly. These results suggested differential contributions of the Robo receptors to the directed elongation of LT and DA3 muscles.

To reveal the relative distribution and expression levels of the Robo receptors in muscles, we took advantage of flies carrying an HA tag that was knocked-in within the genomic region of each of the Robo receptors (Spitzweck et al., 2010), and labeled embryos with anti-HA. Robo1-HA accumulated at the muscle-tendon junction (open arrow) and was observed at low levels on the muscle surfaces (white arrow) (Fig. 2A). Robo2 levels were prominent along ectodermal stripes (open arrow) but not on muscles (Fig. 2B), and Robo3 labeling was indistinguishable from background (Fig. 2C). Double labeling with anti-Robo2 and anti-Stripe (Sr) (Fig. 2D-F) or with CD8-GFP driven by the sr-GAL4 driver (Fig. 2F′-F‴) verified the ectodermal expression of Robo2 at the surfaces of the tendon cells (open arrows in Fig. 2D-F and white arrows in Fig. 2F′-F‴). The relatively high ectodermal expression of Robo2-HA and its lack of staining in the affected LT and DA3 muscles were consistent with a muscle non-autonomous function of Robo2 in mediating LT and DA3 muscle patterning.

To address whether undetectable Robo2 protein was functioning in the muscles, we tested the ability of Robo2 to rescue the robo2 mutant muscle phenotype when expressed in muscles. Driving Robo2 expression in muscles of robo2 mutant embryos using the Mef2-GAL4 driver did not rescue the muscle phenotype, but rather worsened the muscle pattern in all embryos tested as compared with WT embryos overexpressing Robo2 (Fig. 2G-I′). By contrast, knockdown of Robo2 in tendons of robo2^−/+ heterozygous embryos (using sr-GAL4>robo2 RNAi) led to a phenotype similar to that of the robo2 mutant; the LT muscles were mispatterned in 84% of the segments, and the DA3 muscle was aberrant in 31% of the segments (Fig. 2J-L; robo2^+/−,sr>robo2i versus WT, P=0.0001). Importantly, partial rescue of the LT muscle pattern was achieved by expressing Robo2 in the ectoderm using the 69B-GAL4 driver (Fig. 2L; robo2;69B>Robo2 versus robo2, P=0.0001). Moreover, embryos in which robo1 has been inserted into the robo2 locus (Spitzweck et al., 2010) rescued the robo2 muscle phenotype (data not shown), supporting the crucial contribution of Robo2's unique ectodermal distribution rather than its signaling activity.

In summary, the ectodermal expression of Robo2, its knockdown in tendons, and the ectodermal rescue imply that Robo2 functions in the ectoderm to induce the correct muscle pattern.

Next, we addressed the requirement for Slit in the ectodermal activity of Robo2, as associated with guiding the LT and DA3 muscles. We found strong genetic interaction between robo2 and slit in inducing the LT and DA3 muscle pattern (Fig. 3A-C′). Trans-heterozygous embryos for both robo2 and slit exhibited an aberrant pattern of LT muscles in 82% of the segments relative to 0% or 10% in robo2^−/+ or slit^−/+ embryos, respectively (P=0.0001). In the DA3 muscle, 10% of the segments of the trans-heterozygous slit or robo2 embryos exhibited aberrant muscle orientation, in contrast to 0% in either of the single heterozygous mutants (P=0.0001). These results imply that Robo2 activity is mediated through its ligand Slit.

Next, we tested whether Robo2 affects Slit cleavage, which is crucial to short-range Slit signaling (Ordan et al., 2015). We overexpressed Robo2 in the entire embryonic ectoderm using the 69B-GAL4 driver and followed endogenous Slit cleavage by western blotting using an antibody that is reactive to the C-terminus of Slit; this antibody recognizes both full-length Slit (Slit-FL, ∼170 kDa) and the C-terminal cleaved polypeptide (Slit-C, ∼80 kDa). Strikingly, overexpression of Robo2 in the ectoderm led to an average 9.7-fold increase in the ratio of Slit-C relative to Slit-FL levels (P<0.05), implicating Robo2 in promoting Slit cleavage in the ectoderm (Fig. 3D,D′). By contrast, overexpression of Robo2 in the muscles (using the Mef2-GAL4 driver) did not have this effect (not shown). Remarkably, similarly enhanced Slit cleavage was observed following overexpression of truncated Robo2 lacking the cytoplasmic domain (Evans and Bashaw, 2010) (Fig. 3E,E′). Notably, overexpression of Robo1 in the ectoderm similarly enhanced Slit cleavage (data not shown). These results implicated a non-signaling function for Robo2 in promoting Slit cleavage.

We reasoned that covalently bound Slit oligomers, shown to form by previous crystallographic studies (Seiradake et al., 2009), would enable Slit to bind both of the juxtaposing Robo receptors – on the muscle side (Robo1 and/or Robo3) and on the tendon side (Robo2). We therefore tested for Slit oligomerization in embryos. Overexpression of either GFP-tagged cleaved Slit-N (Slit-N-GFP) or uncleavable full-length Slit-UC-myc was induced in the embryonic ectoderm with the 69B-GAL4 driver. The embryo extracts were then boiled under reducing [with β-mercaptoethanol (BME)] or non-reducing (without BME) conditions, separated on SDS-PAGE and analyzed by western blot using anti-GFP or anti-myc. Slit-N-GFP exhibited slower electrophoretic mobility in non-reducing than in reducing conditions (expected to be 124 kDa), supporting the formation of Slit-N-GFP oligomers by disulfide bonds (Fig. 3F, left panel). By contrast, Slit-UC-myc did not exhibit differential electrophoretic mobility in native versus denaturing conditions (Fig. 3F, right panel). This result is consistent with the unique ability of Slit-N to oligomerize.

Taken together, these results suggest that Robo2 in tendons enhances Slit cleavage, producing Slit-N oligomers at the tendon cell membrane that are potentially capable of binding both Robo2 in cis and Robo1/Robo3 in trans, leading to the spatial restriction of Slit signaling to the elongating muscle.

We reasoned that if the primary function of Robo2 is to promote localized Slit-N oligomers, then the knocked-in, membrane-bound, uncleavable Slit (Slit-UC-CD8) would rescue the phenotype of robo2. We therefore recombined knocked-in Slit-UC-CD8 [described by Ordan et al. (2015)] with robo2 and analyzed the phenotype of the DA3 and LT muscles in homozygous embryos. Whereas in robo2 mutants the orientation of the LT muscles was defective in 85% of the segments (Fig. 4A,B), the addition of knocked-in Slit-UC-CD8 showed only 37% defective segments (P=0.0001, n=46; Fig. 4D). Likewise, the D_LT3/D_s ratio was rescued (0.3±0.06 in the robo2;Slit-UC-CD8 embryos relative to 0.16±0.1 in robo2, P=0.015). Given that Slit-UC-CD8 by itself produced a moderate muscle phenotype (Fig. 4C), we concluded that Slit-UC-CD8 is capable of rescuing the robo2 phenotype of the LT muscles (Fig. 4I). For the DA3 muscle, whereas 32% of the segments showed a phenotype in robo2 mutants, robo2 mutant embryos recombined with Slit-UC-CD8 showed a muscle phenotype in only 13% of the segments (P=0.035), and Slit-UC-CD8 mutants showed only 5% defective segments (Fig. 4E-H,J).

Moreover, consistent with a major function of Robo2 in promoting Slit cleavage, overexpression of the active cleaved Slit-N-GFP in tendon cells rescued the robo2 mutant LT muscle phenotype in 73% of the segments (Fig. 4L and Fig. 2L; robo2;sr>Slit-N versus robo2, P=0.0001), whereas expression of Slit-UC-myc alone did not (Fig. 4K and Fig. 2L; robo2;sr>Slit-UC versus robo2, P=0.9268). These results are consistent with a major function for Robo2 in promoting Slit-N accumulation at the tendon cell membrane, presumably by retaining Slit-FL and enhancing its availability for proteolytic cleavage. Slit-N oligomers potentially bind to both Robo2 at the tendon side as well as to Robo1 and/or Robo3 at the muscle side to promote a unidirectional short-range Slit-repulsive signal in muscles (Fig. 4M). In this context the Robo2 cytoplasmic signaling domain is dispensable, consistent with Robo2 function in chordotonal organs (Kraut and Zinn, 2004) and with Robo2 non-autonomous function in the CNS (Evans et al., 2015).

In conclusion, our findings demonstrate a novel, non-signaling role for Robo2 in tendons in regulating the local distribution of active cleaved Slit oligomers at the tendon cell membrane. This signal is crucial in promoting the Slit short-range signaling in muscles that is essential for directional muscle elongation. Such a mechanism might be significant in other setups in which Slit promotes signaling between neighboring cells.

Wild-type flies were of the yw strain. The strains robo2[4], robo2[8], robo2→robo1 (robo1 knocked into robo2), robo3[1], Slit-UC-CD8 (uncleavable membrane-bound Slit/CyO Wg-lacZ), Robo1-HA knock-in, Robo2-HA knock-in, Robo3-HA knock-in and UAS-Robo2-HA were produced by the B. J. Dickson lab (Janelia Research Campus, Ashburn, VA, USA). Mef2-GAL4, 69B-GAL4, sli²/CyO,Yfp robo¹/CyO,Yfp and robo2 RNAi flies were obtained from Bloomington Drosophila Stock Center, Indiana University. Stripe-GAL4 flies were provided by G. Morata (CBMSO, Universidad Autonoma de Madrid, Madrid, Spain). Collier-GFP flies were a gift from Dr Alan Vincent (Centre de Biologie du Développement, UMR 5547 and IFR 109, CNRS and Université Paul Sabatier) (Dubois et al., 2007; Enriquez et al., 2012). Homozygous embryos were recognized by the lack of either CyO,YFP or CyO,Wg-lacZ. UAS-slit-N-GFP and UAS-U-slit-myc and UAS-Robo2-Extra strains were a gift of Greg J. Bashaw (Coleman et al., 2010). Ap-Me580-GFP flies were a gift of Mary Baylies (Folker et al., 2012, 2014). slit² and robo2[8] and Slit uncleavable (Slit-UC) and robo2[8] were recombined by conventional methods.

Mouse anti-Collier (Knot) antibody (1:200) was produced by the Alan Vincent lab (Centre de Biologie du Développement). Additional antibodies included rat anti-Tropomyosin (1:400; Abcam, Ab-50567), goat polyclonal anti-myc (1:1000; Abcam, ab9132), mouse anti-GFP (1:200; Roche, 1184460001), chick anti-β-gal (1:400; Abcam, 9361), chick anti-GFP (1:400; Aves Labs, GFP-1020), guinea pig anti-Stripe (1:400; produced in the T.V. lab), rabbit anti-Robo2 (1:200; produced by B. J. Dickson), chick anti-HA (1:400; Aves Labs, ET-HA100), mouse anti-HA (1:400; Covance, MMs-101p), mouse anti-myc (1:60; Developmental Studies Hybridoma Bank, 9E10) and mouse anti-Slit (1:30; Developmental Studies Hybridoma Bank, The University of Iowa). Secondary fluorescent antibodies were purchased from Jackson Laboratories.

Staged embryos were collected and fixed as previously described (Ashburner et al., 2005). Embryos were visualized with a Zeiss LSM710 confocal microscope and images were processed using Adobe Photoshop.

Staged embryos were dechorionated and balancer was recognized using a fluorescent stereoscope. Then, embryos were placed on a piece of agar glued to a coverslip with 3 M glue, covered with Halocarbon oil 700 (Sigma-Aldrich, H8898), and visualized with a Zeiss LSM710/780 confocal microscope. Images were deconvoluted using AutoQuant (Media Cybernetics, AutoQuant X3) and time series were transformed into a film using ImageJ (NIH).

Protein extract was produced from staged embryos in RIPA buffer. Extract was boiled in sample buffer with or without BME and run on a 7% SDS-PAGE gel, blotted onto nitrocellulose and reacted with mouse anti-Slit antibody (1:1000; Developmental Studies Hybridoma Bank) and HRP-conjugated anti-mouse IgG (1:5000; Jackson ImmunoResearch, 715-035-151). The signal was visualized using SuperSignal (Thermo Scientific). Film was scanned and relative band intensities were measured using ImageJ. Statistical significance was determined by the Mann-Whitney U-test.

We thank A. Vincent, M. Baylies and B. Dickson for gifts of antibodies and/or flies, the Bloomington Stock Center for fly lines, and the Developmental Studies Hybridoma Bank for antibodies. We also thank N. Konstantin for corrections to the manuscript.