Netrins guide migration of distinct glial cells in the Drosophila embryo

During animal development, cells often change shape and position by morphogenetic movements as well as by active migration. Precise coordination of these processes during organogenesis is achieved by extensive cross-talk between motile cells and their environment. The most complex organ, the nervous system, is composed of a variety of neuronal cell types that become connected to each other and to non-neuronal cells in a highly specific manner to form functional networks. The accuracy with which these cells extend their neurites, read environmental signals and finally connect to their appropriate target sites needs to be controlled by an intricate molecular machinery (reviewed by Dickson, 2002; Grunwald and Klein, 2002; Huber et al., 2003; Chilton, 2006).

Among the factors involved are Netrins, secreted molecules that form a concentration gradient that is interpreted by navigating axons (reviewed by Livesey, 1999; Barallobre et al., 2005). Netrins were first discovered in vertebrates as neurotrophic morphogens that are secreted from the floor plate and attract neurons to cross the midline (Tessier-Lavigne et al., 1988; Placzek et al., 1990). The two vertebrate Netrins (netrin 1 and netrin 2) are highly homologous to the Caenorhabditis elegans Uncoordinated-6 (UNC-6) protein (Ishii et al., 1992), which was discovered together with UNC-5 and UNC-40 as guidance cues for pioneer axons (Hedgecock et al., 1990). It was shown that UNC-6 and the vertebrate Netrins primarily serve as diffusible ligands (Kennedy et al., 1994). UNC-5 and UNC-40 [the vertebrate homologues are Unc5 and deleted in colorectal carcinoma (Dcc), respectively] are Netrin receptors that promote either axon repulsion (UNC-5/Unc5) (Leung-Hagesteijn et al., 1992; Leonardo et al., 1997) or attraction (UNC-40/Dcc) (Chan et al., 1996; Keino-Masu et al., 1996).

In Drosophila, two Netrins (NetA and NetB) are described as guidance cues, e.g. for commissural and longitudinal axons and motoneurons (Harris et al., 1996; Mitchell et al., 1996; Labrador et al., 2005). Attraction of commissural axons towards the midline is mediated by the receptor Frazzled (Fra), the Drosophila homologue of UNC-40/Dcc (Kolodziej et al., 1996). Fra seems to guide commissural axons both by cell-autonomous signalling events (Forsthoefel et al., 2005; Brankatschk and Dickson, 2006; Garbe and Bashaw, 2007) and by capturing Netrin and presenting it to other neurons (Hiramoto et al., 2000). Repulsion of motoaxons away from the source of Netrin expression, however, is mediated via Unc5 (Keleman and Dickson, 2001; Labrador et al., 2005).

Both Netrins are secreted by cells of the ventral midline, a structure comparable to the vertebrate floor plate. Glial cells of the ventral midline not only secrete diffusible molecules that act as chemoattractants or repellents for navigating neurons (reviewed by Jacobs, 2000), but also serve as guidepost cells for axons projecting along or across the midline (Klämbt et al., 1991) (reviewed by Auld, 1999; Tear, 1999; Jacobs, 2000; Chotard and Salecker, 2004; Parker and Auld, 2004). Besides midline glia, other populations of glial cells in the developing nervous system interact with neurons. Longitudinal glia (LG), the embryonic interface glia, are involved in both the guidance of pioneer neurons and the fasciculation of longitudinal axon bundles (Hidalgo et al., 1995; Hidalgo and Booth, 2000) (reviewed by Auld, 1999; Parker and Auld, 2004; Parker and Auld, 2006). Glial cells in the transition zone between the central and peripheral nervous systems (the CNS and PNS, respectively) guide both sensory neurons into the CNS and motoneurons out into the periphery (Sepp et al., 2000; Sepp et al., 2001; Parker and Auld, 2004).

All glial cells are themselves highly motile and migrate over considerable distances to occupy characteristic positions in the developing (embryonic) nervous system (reviewed by Parker and Auld, 2004; Klämbt, 2009). How do glial cells navigate? Do they use similar molecules and mechanisms to axons? In order to understand the mechanisms involved in glial cell migration and guidance, we analysed the influence of Netrins and their receptors on embryonic glial cells. We previously described the stereotypic migration of embryonic peripheral glia (ePG) as well as a collection of cell-specific markers to identify single, or subsets of, glial cells in the CNS and PNS of Drosophila (Beckervordersandforth et al., 2008; von Hilchen et al., 2008). This enabled us to analyse the effects of loss-of-function mutations of both Netrins and their receptors Unc5 and Fra in glial cells at single-cell resolution. Here we show that in Netrin mutants, two distinct populations of glial cells are affected. LG fail to migrate medially in the early stages of neurogenesis and we provide evidence that early Netrin-dependent LG guidance requires the expression of Fra already in the precursor cell, the longitudinal glioblast (LGB). By contrast, later in development, nearly all ePG transiently express unc5 (unc-5 – FlyBase) mRNA during their migratory phase. In unc5 mutants, however, two of these cells in particular show defective migration and stall at, or close to, their place of birth in the CNS. Both phenotypes are reversible by cell-specific rescue experiments, indicating that Netrin-mediated signalling via Fra (in LG) or Unc5 (in ePG) is a cell-autonomous effect. Furthermore, we demonstrate that the two Netrins are expressed within the neuroectoderm and in neuroblasts between stages 10 and 11 and can redundantly guide the LGB. Our results also suggest an exclusive role for NetB in mediating the repulsion of ePG away from the ventral midline.

The following fly strains were used: wild type OregonR; NetA^Δ, NetB^Δ, NetAB^Δ, NetA^Δ-NetB^myc-TM (Brankatschk and Dickson, 2006); fra³ and fra⁴ [Bloomington Drosophila Stock Center (BDSC)]; unc5⁸ (Labrador et al., 2005); sim² (BDSC); cas-Gal4 (Hitier et al., 2001); elav-Gal4 (BDSC); gcm-Gal4 combined with UAS-ncGFP (gift from A. Giangrande, IGBMC Strasbourg, France); Mz605-Gal4 and Mz1580-Gal4 (Ito et al., 1995); pros-Gal4 C21 (gift from F. Matsuzaki, CBD Kobe, Japan); repo-Gal4 (Sepp et al., 2001); sca-Gal4 (Klaes et al., 1994); sim-Gal4 (gift from C. Klämbt, University of Münster, Germany); UAS-NetA and UAS-NetB (Mitchell et al., 1996); UAS-fra (Kolodziej et al., 1996); UAS-HA-unc5 (Keleman and Dickson, 2001); UAS-nlacZ (BDSC); UAS-Rho1.V14 and UAS-Rho1.N19 (BDSC). For labelling of distinct ePG and for rescue experiments the UAS-Gal4 system was used (Brand and Perrimon, 1993). Whenever required, lacZ- or GFP-tagged balancer chromosomes (BDSC) were used for identification of genotypes.

Embryos were fixed as described (Rogulja-Ortmann et al., 2007). Primary antibodies were mouse anti-Fas2 (1D4, 1:10) and mouse anti-Pros (1:5), both from DSHB (Iowa city, IA, USA); rabbit anti-Repo (1:500) (Halter et al., 1995); rabbit anti-Msh (1:50; gift from C. Doe, University of Oregon, USA); guinea pig anti-Nazgul (1:500); chicken anti-β-gal (1:1000; Abcam, Cambridge, UK); and rabbit anti-Fra (1:500; gift from A. Stollewerk, University of London, UK). Cy3-, Cy5- or FITC-conjugated donkey secondary antibodies were used (1:250; Jackson ImmunoResearch, UK). For imaging, a Leica TCS SPII confocal microscope was used. Images were processed using Leica Confocal Software and Adobe Photoshop.

In vitro transcription and labelling of RNA were performed using the Dig-RNA labelling mix according to the manufacturer’s instructions (Roche Diagnostics, Mannheim, Germany). In situ hybridisation in combination with anti-Repo antibody staining was performed as described (Altenhein et al., 2006). Images were taken with a Zeiss Axioskop 2 microscope and processed using AxioVision Software and Adobe Photoshop.

To investigate whether Netrins guide glial cell migration, we analysed mutants in which both Netrins are removed (NetAB^Δ) (Brankatschk and Dickson, 2006). Embryos at late stage 16 were stained for the glial marker Reversed polarity (Repo) and the neuronal cell adhesion molecule Fasciclin 2 (Fas2). We analysed the pattern of ePG in abdominal segments. Nearly all of the 12 ePG in each hemisegment (hs) have a unique identity, a stereotypic migration and an invariant final position at the end of embryogenesis (for details, see von Hilchen et al., 2008). In wild type, six of these cells finally align along the intersegmental and segmental nerves (ISN and SN, respectively) distal to the nerve branch of the SNc (Fig. 1A-A′, arrows). In NetAB^Δ mutants (Fig. 1B-B′), the number of ePG distal to the SNc was reduced in 94% of hs [NetAB^Δ, median of four cells (n=68 hs); wild type, median of six cells (n=100 hs); Fig. 1E]. Sometimes, more than three ePG were detected in the exit area of affected hs (between the SNc and the CNS), suggesting that the ePG missing distal to the SNc stall in this area (Fig. 1B,B′, asterisk). This phenotype also occurred in NetB^Δ single mutants at nearly the same penetrance, but with a slightly decreased expressivity (Fig. 1E). NetA^Δ single mutants never showed this phenotype, indicating that ePG guidance is mediated by NetB, although the reduced expressivity in NetB^Δ mutants compared with NetAB^Δ suggests a synergistic effect and hence some functional redundancy between the two Netrins.

In addition to the stalling phenotype of ePG, 29% of hs of NetAB^Δ mutants showed ectopic glial cell clusters in the dorsal periphery in close proximity to ePG11 (Fig. 1B,B′, curved arrow). Although all NetAB^Δ mutant embryos showed ectopic clusters, the number was variable, ranging from one to ten clusters per embryo. They were not obviously associated with neuronal structures. Analysis of the CNS of corresponding hs revealed a lack of cells dorsal to the longitudinal connectives, which are referred to as interface glia or LG. Analysis of Netrin single mutants revealed ectopic clusters with reduced penetrance and expressivity mainly in the posterior segments of NetA^Δ mutants, but never in NetB^Δ mutants (Fig. 1E). Furthermore, longitudinal axon tracts displayed fasciculation defects, especially in hs in which LG were missing (Fig. 1B). These longitudinal axon defects have been observed previously, but without reporting a glial phenotype (Harris et al., 1996; Mitchell et al., 1996).

We further investigated the two Netrin receptors Unc5 and Fra. In unc5⁸ mutants, ePG stalling occurred with nearly identical expressivity as in NetAB^Δ, but we never found ectopic LG clusters (Fig. 1C,E); these were detectable only in fra³/fra⁴ mutants in ∼18% of hs (Fig. 1D,E, curved arrows). In addition to ectopic LG clusters, fra³/fra⁴ mutants also showed a mild ePG stalling phenotype, with mainly one cell stalling in the exit area in 49% of hs (Fig. 1D′,E).

All LG are progeny of the LGB, which arises at the lateral edge of the neuroectoderm at stage 10. It successively divides while migrating towards the ventral midline, and gives rise to nine cells that align in rows dorsal to the longitudinal connectives (Jacobs et al., 1989; Schmidt et al., 1997) (reviewed by Auld, 1999; Parker and Auld, 2004; Stacey et al., 2007). In order to prove that ectopic clusters of glial cells in the periphery of NetAB^Δ, NetA^Δ and fra³/fra⁴ mutants indeed consist of LG, we stained for LG-specific markers. All LG can be labelled with an antibody against the Muscle segment homeobox (Msh; Drop – FlyBase) protein (Fig. 2A-B′, yellow arrowheads) from stage 11 onwards (Shishido et al., 1997; Hidalgo et al., 2001; Beckervordersandforth et al., 2008). In addition, the medial intersegmental nerve root glia (M-ISNG) is Msh positive (Fig. 2B′,B′, yellow arrows) (Beckervordersandforth et al., 2008). In hs showing no ectopic LG, we detected nine plus one Msh-positive cells (LG and M-ISNG, n=100 hs) in the CNS (Fig. 2F-F′, hs A3). In the majority of affected hs (>75%), nine LG were detected within ectopic clusters in the PNS (Fig. 2F, curved arrow), while the Msh-positive M-ISNG was still present in the CNS (Fig. 2F′,F′, yellow arrows).

We also used an antibody directed against a protein, which we termed Nazgul (Naz), that is exclusively expressed in LG and cell body glia (CBG) (Fig. 2C). Naz is a cytoplasmic protein that is expressed in LG and CBG from stage 12 onwards until the end of embryogenesis (our unpublished data). Staining NetAB^Δ mutant embryos with anti-Naz unambiguously identified the ectopic clusters as mispositioned LG (Fig. 2G, curved arrow), leaving clear gaps in Naz expression (and hence gaps of LG) on longitudinal axons in corresponding hs (Fig. 2G, dotted circle). Exceptions to this phenotype included fewer cells within ectopic clusters or scattered ectopic LG (see Fig. S1 in the supplementary material).

Having shown that most ectopic clusters consist of the entire LGB lineage, we next addressed whether expression of the transcription factor Prospero (Pros) also occurs within ectopic clusters. In wild-type stage 16 embryos, Pros is differentially expressed in six anterior LG (Fig. 2D) (Doe et al., 1991; Griffiths and Hidalgo, 2004; Beckervordersandforth et al., 2008). In NetAB^Δ mutants, on average six LG also expressed Pros (Fig. 2H) (although the standard deviation was higher than for the wild type). These data indicate that not only is cell number normal, but also that lineage-dependent cell specification of LG occurs normally in ectopic clusters and hence is independent of their normal position and postulated interactions with longitudinal axons (Griffiths et al., 2007). We also stained for the LG-specific markers in NetA^Δ and fra³/fra⁴ mutants to confirm that the ectopic clusters in these mutants consist of LG (not shown).

The fact that the LG phenotype occurs already at early stages, immediately after the LGB has divided once, and that most ectopic clusters consist of all nine LG, strongly suggest that Netrin-Fra interaction is required already at late stage 10/early stage 11. We asked whether Fra acts cell-autonomously within the LGB lineage. In agreement with previous data (Kolodziej et al., 1996), we could not detect any Fra protein in glial cells, including the LGB and its early progeny. Both anti-Fra antibody staining and in situ hybridisation revealed a broad fra expression domain lateral to the ventral neurogenic region at stage 10 (Fig. 3A). The LGB delaminates from this region, and in some cases we were able to detect fra mRNA within the Repo-positive LGB at early stage 11 (Fig. 3B). From these findings, we assume that Fra acts already in the LGB and becomes downregulated prior to, or immediately after, delamination. If such early Fra expression is required and sufficient to guide LG medially, it should be possible to rescue the cluster phenotype of fra³/fra⁴ mutants with an early glial-specific Gal4 driver. Indeed, gcm-Gal4-driven expression of UAS-fra in a fra mutant background was able to restore the phenotype (Fig. 3D and Table 1), whereas repo-Gal4-driven expression of UAS-fra could not (Table 1). We tested further Gal4 drivers that are known to be expressed in the LGB lineage, but only Mz1580 was able to weakly restore the LG phenotype (Table 1). Neuronal expression of UAS-fra (e.g. driven by elav-Gal4) did not rescue this phenotype at all (Table 1). Additionally, none of the glial Gal4 drivers rescued the mild ePG phenotype of fra mutants. These data indicate that Fra acts cell-autonomously within the LGB lineage, but has a non-autonomous effect on ePG migration. The stalling of mainly one ePG per hs in fra³/fra⁴ mutants is most likely a secondary consequence of pathfinding defects of peripheral axons.

Fra is known to promote an attractive signal, e.g. for navigating neurons towards the ventral midline, where both Netrins are expressed. In order to test whether Netrins supplied by cells of the ventral midline are responsible for LG guidance, we used the sim-Gal4 driver to rescue the LG phenotype of NetAB^Δ mutants. To our surprise, sim-Gal4-driven expression of either UAS-NetA or UAS-NetB only weakly restored the LG phenotype (Table 1). To further analyse the role of midline-derived Netrins in LG guidance, we investigated single minded (sim²) mutants. Sim is a key regulator of midline differentiation, and in sim² mutants midline cells are completely absent, leading to severe axon pathfinding defects and to a collapse of longitudinal axon tracts. Longitudinal glia, however, were positioned medially in sim² mutants and lay on top of collapsed longitudinal axons (Fig. 4B,D). Expression of the LG-specific markers Naz and Pros appeared unchanged in sim² mutants, as compared with wild type (Fig. 4A-D), but owing to the severity of the phenotype exact cell counts were not possible. We never found ectopic clusters of LG in the periphery of sim² mutants though, which we would expect if the midline served as a Netrin source for LG guidance.

Thus, we investigated the expression of both Netrins in early stages by mRNA in situ hybridisation (Fig. 4E-F′). Between stages 9 and 11, Netrin expression is highly dynamic in different tissues, including mesoderm, neuroectoderm and neuroblasts (NBs). We used additional markers to define more precisely which NBs express Netrins (see Fig. S2 in the supplementary material). Some of these expression domains that included Netrin-expressing NBs lay medial to the site where the LGB delaminates (Fig. 4E-F′, arrowheads) and could therefore potentially guide the LGB and its early progeny. We observed differences between NetA and NetB expression: in the posterior segments A5 to A8, NetA was predominantly expressed, whereas NetB expression was weak here. These differences in expression might explain why ectopic LG clusters occur in NetA^Δ single mutants only in posterior segments, whereas NetB^Δ single mutants never show ectopic clusters (Fig. 1E).

To prove that early Netrin expression within the ventral neurogenic region and in NBs serves as a cue for LG guidance, we expressed either UAS-NetA or UAS-NetB under the control of scabrous (sca)-Gal4, which drives expression broadly within the neuroectoderm from stage 8/9 onwards. In NetAB^Δ mutants, sca-Gal4-driven expression of either Netrin resulted in complete rescue of the cluster phenotype in ∼90% of embryos (Table 1). Ectopic expression of either Netrin with sca-Gal4 in an otherwise wild-type background also occasionally resulted in ectopic clusters. Why is it that sca-driven expression of Netrins restores the LG phenotype of NetAB^Δ, but induces the same phenotype in an otherwise wild-type background? In the absence of endogenous Netrin, the sca-driven expression of either UAS-NetA or UAS-NetB is sufficient to significantly rescue LG guidance. However, in the presence of endogenous Netrin expression, additional Netrin expression by sca-Gal4 might overflow a medial-to-lateral gradient and hence abolish the directionality of the signal. This would support the hypothesis that Netrin-mediated signalling in LG serves as an instructive cue that provides directionality to LG migration, rather than serving as a permissive signal that provides motility per se without any direction. In order to prove this model, we ectopically expressed either Netrin in an otherwise wild-type background at early stages in the vicinity of the LGB, as induced by Krüppel-Gal4 within its gap-gene domain, or in a striped pattern in each segment, as induced by engrailed-Gal4. All three drivers resulted in ectopic clusters of LG (with variable expressivity and penetrance; for details, see Fig. S3 in the supplementary material), which further supports the proposal that both Netrins act as instructive cues for LG migration.

It was previously reported that unc5 is expressed in ePG (Keleman and Dickson, 2001; Freeman et al., 2003), although the resolution was insufficient to distinguish which. We observed transient expression of unc5 mRNA in ePG between stages 13 and 16 (Fig. 5A,B). But which ePG is affected in unc5⁸ mutants? With the capacity to identify nearly all ePG with molecular markers and by position at stage 16 (von Hilchen et al., 2008), we found that ePG6 and ePG8, which are progeny of NB2-5, were missing distal to the SNc in unc5⁸ (and NetAB^Δ) mutants. cas-Gal4>UAS-nlacZ was used to selectively label ePG6 and ePG8 (Fig. 5C, arrowheads) and confirmed that in homozygous unc5⁸ mutants, these two ePG either stall in the exit area proximal to the SNc (Fig. 5D, arrowheads) or remain in the CNS (Fig. 5D′,E; see Fig. S4 in the supplementary material). Occasionally, ePG5 also stalled in the exit area (Fig. 5D, asterisk). We used an antibody (anti-SP2637) that is specific to a subset of ePG to further validate the selectivity of the observed phenotype in unc5⁸ mutants (see Fig. S4 in the supplementary material).

In order to examine whether Unc5 function is required cell-autonomously in glial cells, rescue experiments were performed with either cas-Gal4 or repo-Gal4. cas-Gal4-driven expression of unc5 in homozygous unc5⁸ mutants led to significant rescue of the phenotype. The normal number of six ePG distal to the SNc was restored in 49% of hs (compared with 6% in unc5⁸ mutants) and five ePG were found distal to the SNc in 39% of hs (compared with 24% in unc5⁸ mutants) (Fig. 6B,F). Rescue experiments with repo-Gal4 restricted the expression of UAS-unc5 to glial cells and also resulted in a significant rescue of the ePG stalling phenotype (Fig. 6D,F), with the normal number of six ePG distal to the SNc in 49% of hs (compared with only 6% in unc5⁸ mutants) and five ePG distal to the SNc in 33% of hs (compared with 24% in unc5⁸ mutants). Sometimes, one of the three proximal ePG (ePG1-3) migrated further distal, resulting in seven ePG lying distal to the SNc (11%, Fig. 5F). Although pan-glial expression of unc5 in an otherwise wild-type background does not affect ePG migration, LG are often shifted to more lateral positions within the CNS (Freeman et al., 2003). We frequently observed two or more LG in the periphery [penetrance of 76.4% (n=17 embryos) and expressivity of 19.5% (n=104 hs)] and confirmed the identity of these ectopically located LG by anti-Naz staining (Fig. 6C,E, yellow arrowheads).

The finding that only NetB^Δ single mutants show an ePG stalling phenotype and that Unc5 is required for ePG guidance suggest that ePG are repelled away from the CNS into the periphery by NetB. To further illustrate this, we used sim-Gal4 to drive either UAS-NetA or UAS-NetB in midline cells of NetAB^Δ mutants. We observed a clear rescue of the ePG stalling phenotype with UAS-NetB (Fig. 7C,E), but not with UAS-NetA (Fig. 7B,E). To prove that the midline serves as an endogenous NetB source, we examined sim² mutants with respect to ePG migration. Peripheral nerves were severely affected in sim² mutants (Fig. 7D, asterisk), which hampers the analysis. Nevertheless, ePG were clearly missing in the PNS (Fig. 7D, arrows), which we further confirmed by anti-SP2637 staining (see Fig. S4 in the supplementary material). These data reveal that NetB secreted from the ventral midline serves as a guidance cue for ePG6 and ePG8.

Based on the present data, we postulate a dual role for Netrin-mediated signalling in glial cell migration. According to our model, early in neurogenesis, Netrins guide the LGB and its progeny from the lateral edge of the neuroectoderm towards a medial position. We show that at these early stages, Netrins are expressed by cells of the neuroectoderm as well as by NBs, and that this Netrin source most likely attracts the LGB via Fra. Ectopic expression of Netrins in the vicinity of the LGB might abolish a possible ventral-to-dorsal gradient and hence (occasionally) results in ectopic clusters (see Fig. S3 in the supplementary material). Additionally, we attempted to express Netrins in the dorsal area of the embryo and thereby attract the LGB and its progeny in the wrong direction. Unfortunately, none of the tested drivers showed Gal4 expression at the appropriate stage and intensity. Further experiments are needed to prove this model.

The LGB delaminates from the lateral neuroectoderm close to the sensory organ precursor-derived ePG11 (60-65% dorsoventral axis, where 0% is the ventral midline). In wild type, it migrates medially while proliferating, whereas we believe that in Netrin and fra mutants the LGB remains (and proliferates) at its place of birth and does not migrate at all in affected hs. Morphogenetic movements during germ band retraction, mesoderm migration and dorsal expansion of the epidermis complicate this issue. Nevertheless, ectopic clusters mainly remain in close proximity to ePG11. Although ectopic LG have no contact to axons, the lineage develops normally with respect to cell number and marker gene (Msh, Naz, Pros) expression. This is contrary to published data on the development and differentiation of LG, which have been postulated to depend on an intimate interaction with longitudinal axons (Griffiths and Hidalgo, 2004; Griffiths et al., 2007). Nevertheless, LG play an important role in the navigation and fasciculation of longitudinal axons (Hidalgo et al., 1995). Accordingly, in hs of Netrin and fra mutants, in which LG are mispositioned from the earliest stages, we found that the longitudinal axon tracts are thinner, show aberrant projections and fasciculation defects (see Fig. S5 in the supplementary material). These neuronal phenotypes were reported previously, but without noticing that the LG are missing in these hs (Harris et al., 1996; Kolodziej et al., 1996; Mitchell et al., 1996). Similar neuronal phenotypes can be induced by ectopic expression of unc5 in glial cells (repo>unc5) (see Fig. S5 in the supplementary material), which only affects LG and shifts them away from the midline to more lateral positions or even into the PNS. In some hs with ectopic LG clusters, longitudinal tracts show a weaker phenotype. In these cases, other glial cells (from within the same hs or an adjacent hs) fill the gaps on top of the longitudinal axons and hence seem to compensate for the loss of LG (data not shown; E. Giniger, personal communication).

From these data, we conclude that the longitudinal axon phenotypes observed in Netrin and fra mutants are, at least partially, a secondary effect of the lack of LG in corresponding neuromeres. Additionally (and somewhat confusingly), these neuronal phenotypes in fra mutants can be partially rescued by elav-Gal4 and Mz605-Gal4 (Hiramoto et al., 2000; Keleman and Dickson, 2001), but neither rescues the LG phenotype (Table 1). Further experiments with other Gal4 drivers that allow a more restricted spatio-temporal expression of UAS-fra might help to resolve this issue.

The second population of glial cells that is guided by Netrin-mediated signalling comprises ePG. Nine ePG migrate from the ventral nerve cord into the PNS of each abdominal hs, but it is ePG6 and ePG8 in particular, both progeny of NB2-5, that show a stalling phenotype in NetAB^Δ, NetB^Δ and unc5⁸ mutants. We show by rescue experiments and analysis of Netrin single mutants that only NetB provided by cells of the ventral midline repels ePG6 and ePG8 via the Unc5 receptor. Although both Netrins are expressed by the ventral midline, they clearly do not share a redundant function for ePG guidance. A similar observation has been reported for unc5-expressing motoaxons, which respond differently to each Netrin (Mitchell et al., 1996; Winberg et al., 1998). To date, the nature of these differences between the two Netrins in combination with Unc5 remains unresolved. In NetA^Δ-NetB™, only NetB is expressed, but is tethered to the membrane of the cell. In these embryos, no ePG stalling was observed (data not shown), further supporting the notion that only NetB is required for ePG migration and indicating that this signalling is at short range. But why do all ventrally derived ePG express unc5 mRNA transiently during their migratory phase? Further work will be needed to clarify why NetB-Unc5 signalling is selectively required for normal migration of ePG6 and ePG8.

It is widely accepted that embryonic glial cells use neurons or neuronal processes as the substrate for migration (reviewed by Cafferty and Auld, 2007). It was shown previously that most migrating ePG follow certain axonal projections (reviewed by Silies et al., 2007). Could such neuron-glia contact be sufficient for proper guidance? The questions would then be (1) how do glial cells actually identify their respective neuronal projections and (2) how is directionality of migration given? Four-dimensional analysis of ePG migration, however, has revealed that ePG6 and ePG8 do not necessarily follow axons, but may also use other glial cells as substrate (von Hilchen et al., 2008). These two cells leave the CNS later than other ePG, they are the only ePG that can overtake other cells, and they may migrate on top of ePG rather than along peripheral nerves. Is this possible lack of axonal association (and perhaps adhesion) the reason why these cells need an additional guidance system?

The initial migration of the LGB and its early progeny cannot occur along axons because at these early stages axonal projections are not yet established. As discussed, in most cases the LG phenotype affects the entire LGB lineage and hence is an early guidance defect. So it might well be that early LGB guidance is dependent on Netrin-Fra signalling without any neuronal contribution. After the first division of the LGB, neuronal projections are established, Net-Fra attraction is no longer required and fra expression is switched off. The results of our fra rescue experiments show that at least the timing of fra expression is crucial: gcm-Gal4-induced fra expression can rescue the LG phenotype, whereas a slightly later expression driven by repo-Gal4 cannot. The expressivity of the LG phenotype in NetAB^Δ mutants is only 30%. How are 70% of the LGB ‘rescued’? A second, as yet unknown, signalling mechanism could guide the LGB medially, either by ventral attraction or dorsal repulsion. Similar redundancies have been reported, e.g. for border cell migration in the ovary or germ cell migration in early embryogenesis (reviewed by Ribeiro et al., 2003).

In addition to the selectivity of the phenotypes in different populations of glial cells, as discussed above, another interesting observation comes from our rescue experiments in unc5⁸ and fra³/fra⁴ mutants. We performed control experiments to test whether pan-glial expression of either receptor affects glial cell migration. Glial expression of UAS-fra in an otherwise wild-type background does not alter the glial pattern in the CNS or PNS. Since unc5 is expressed normally in these experiments, repulsion of ePG6 and ePG8 into the periphery occurs as normal. By contrast, pan-glial expression of unc5 is able to shift LG to a more lateral position in the CNS [as reported previously (Freeman et al., 2003)] and LG can even leave the CNS and lie in the periphery, whereas all other glial cells are properly positioned. Why do only certain glial cells react upon Netrin-mediated signalling? More precisely, what conveys the ability for LG to ‘read’ Netrin-mediated signalling? In addition to the receptors, cells might require downstream molecules that could be differentially expressed and hence provide competence to react to Netrins. Several such molecules have been described for both vertebrates and invertebrates (Forsthoefel et al., 2005; Li et al., 2004; Li et al., 2006; Lucanic et al., 2006; Shekarabi et al., 2005). We have analysed loss-of-function mutants for Drosophila homologues of these possible downstream molecules, but none showed phenotypes comparable to those of fra³/fra⁴ or unc5⁸ (data not shown). Previous data demonstrate a function for the small GTPases Rac1 and Rho1 in ePG migration in Drosophila (Sepp and Auld, 2003), and it was recently shown that they can act downstream of Unc5 signalling in vertebrates (Picard et al., 2009). We expressed a dominant-negative form of Rho1 using cas-Gal4 in an otherwise wild-type background and obtained stalling of ePG6 and ePG8 (cas>Rho1N19, data not shown). Expression of a constitutively active form of Rho1 in ePG6 and ePG8 in an unc5 mutant background (unc5⁸; cas>Rho1V14), however, did not restore their stalling phenotype. Although we cannot rule out the possibility that ectopic expression of such constructs leads to artificial phenotypes, these data indicate that Rho1 is not downstream of Unc5, but rather acts in parallel. Further experiments will be needed to unravel the signalling complex of Unc5 and Fra/Dcc, and glial cell migration in the Drosophila embryo might serve as a powerful model system for this purpose.

We thank E. Giniger for flies, unpublished data, helpful discussions and comments on the manuscript; A. Giangrande and I. Salecker for fruitful discussions; M. Brankatschk, B. Dickson, C. Doe, T. Evans, C. Klämbt, F. Matsuzaki, A. Mauss, K. Saigo and A. Stollewerk for fly stocks and antibodies; E. Griebel for help with statistical calculations; O. Vef, B. Groh-Reichert, J. Dietrich and J. Seibert for technical assistance; and C. Rickert for initial help with the phenotype. This work was supported by grants from the Deutsche Forschungsgemeinschaft to B.A. and G.M.T.