Axonal wrapping in the Drosophila PNS is controlled by glia-derived neuregulin homolog Vein

Neuronal function requires the transmission of electric signals. It has long been known that glial cells insulate axons to ensure fast and reliable transmission of action potentials. In vertebrates, fast saltatory conductance is possible due to myelin-forming glial cells. Long segments of the axons are tightly insulated by several glial wraps, leaving free only the nodes of Ranvier (Nave and Trapp, 2008; Poliak and Peles, 2003; Sherman and Brophy, 2005). In addition, glial cells support the integrity of the axons (Fünfschilling et al., 2012; Nave, 2010). In the mammalian peripheral nervous system (PNS), the activation of the EGF receptor by axonally provided neuregulin 1 (Nrg1) type III ensures the correct differentiation of Schwann cells (Michailov et al., 2004; Taveggia et al., 2005). However, following injury, remyelination is not controlled by axonally derived Nrg1. Instead, denervated Schwann cells transiently express Nrg1 type I to promote Schwann cell differentiation and remyelination in an autocrine/paracrine manner (Stassart et al., 2013).

In Drosophila, the organization of the segmentally organized peripheral nerves is remarkably conserved compared with vertebrates. In all abdominal segments, axons of 42-44 sensory and 34-36 motoneurons connect the body wall with the central nervous system (CNS) (Bodmer and Jan, 1987; Bodmer et al., 1989; Landgraf and Thor, 2006; Mahr and Aberle, 2006; Merritt and Whitington, 1995; Sink and Whitington, 1991). The different abdominal nerves (A1-A8) vary in length. The A2 nerve reaches ∼500 µm in length and is thus six times shorter than the A8 nerve, which is ∼3000 µm in length. Sensory and motoaxons are accompanied by three different glial cell types: perineurial glia, subperineurial glia (SPG) and wrapping glia (Stork et al., 2008).

The wrapping glial cells of Drosophila resemble the non-myelinating Schwann cells that form the so-called Remak fibers in mammals (Nave and Salzer, 2006; Rodrigues et al., 2011). The wrapping glial cells do not divide during larval stages but rather grow to a considerable size. In the third larval instar stage, the wrapping glial cells have engulfed all axons but never form multiple membrane wraps as observed in the vertebrate myelin (Stork et al., 2008). The lineage and identity of all peripheral glial cells found in the larval PNS has been recently described (von Hilchen et al., 2013). Along every abdominal nerve, only three to four wrapping glial cells are found. The SPG is also generated during early embryogenesis. These cells do not divide once they are born and persist until adult stages (Awasaki et al., 2008; Stork et al., 2008; Unhavaithaya and Orr-Weaver, 2012; von Hilchen et al., 2013). The SPG forms the blood-brain barrier by establishing intensive cell-cell contacts characterized by pleated septate junctions (pSJ). These cell junctions are structurally and molecularly related to the septate-like junctions at the paranodes in vertebrates (Banerjee and Bhat, 2007; Banerjee et al., 2006; Baumgartner et al., 1996; Peles et al., 1997; Stork et al., 2008). In all abdominal nerves, only four to five subperineurial glial cells can be identified. The perineurial glial cells, the function of which is currently unknown, cover all nerves. These cells are able to divide throughout development and their number reflects the length of the respective peripheral nerve. How this number is adjusted and which mechanisms orchestrate the differentiation of wrapping and subperineurial glial cells is currently unknown.

In the present study, we first provide a thorough transmission electron microscopy (TEM) analysis of the developing peripheral nerves. The wrapping glia stretches out during larval stages to reach a length of up to 1 mm. We then show that wrapping glial cell differentiation is controlled by activation of the EGF receptor through the Drosophila neuregulin homolog encoded by vein. Furthermore, cell type-specific knockdown experiments demonstrate a cell-autonomous requirement of Vein during glial differentiation. Vein is not required in neurons but rather in wrapping glial cells in order to induce efficient axonal wrapping. This resembles the mode of EGF receptor activation seen during regenerative myelination in the mammalian nervous system and sheds light on the evolution of glial differentiation.

The relatively constant number of wrapping and subperineurial glial cells in each abdominal nerve in relation to the enormous growth of the larva points towards an extreme hypertrophic growth and differentiation potential of these cells (von Hilchen et al., 2013; and see Fig. 1). To address the molecular mechanisms underlying the differentiation of the wrapping glia, we first performed a careful ultrastructural analysis of the abdominal nerves. The exact differentiation of the wrapping glial cells can only be analyzed by electron microscopy. In order to obtain quantitative data on the nerve ultrastructure, analysis has to occur on identified nerves at stereotyped positions (Fig. 1A). Therefore, we developed a whole-mount embedding procedure, which enables us to determine the identity of each nerve and the exact position along the nerve (supplementary material Fig. S1; and see Materials and Methods). In general, nerves A1 to A7 appeared very similar. We counted an average of 78 axon profiles in each of these nerves (n=47 nerves), which corresponds well to the previously counted numbers of moto- and sensory neurons (Bodmer and Jan, 1987; Bodmer et al., 1989; Landgraf and Thor, 2006; Mahr and Aberle, 2006; Merritt and Whitington, 1995; Sink and Whitington, 1991). By contrast, the A8 nerve is different, as it is made up by ∼110 axons (n=18 nerves).

To determine the extent of glial wrapping of axons, we defined a wrapping index, which indicates the percentage of individually wrapped axons or axon bundles in relation to the total number of axons (see Material and Methods). Our ultrastructural analysis demonstrates that, in wild type, the wrapping index is ∼20%. The wrapping is relatively constant at 2 µm, 50 µm, 200 µm or 400 µm distance from the CNS, suggesting that, at third-instar stage, the wrapping of axons occurs independently of the position along the anterior-posterior axis (Fig. 1A; supplementary material Fig. S1).

Interestingly, we found that tracheal cells are located within the nerve close to the CNS but are not found in the nerve 200 µm or more away from the CNS (supplementary material Fig. S1D,E, arrows). Tracheal cells establish extensive autocellular pSJ; however, they do not form pSJ with the abutting subperineurial glial cells, which also form autocellular pSJ (supplementary material Fig. S2A-C). Tracheal cells are always found in the perineurial glial layer, often covered by processes of the SPG (supplementary material Fig. S2A, green shading). The SPG forms spot-adherens junctions with the tracheal cells (supplementary material Fig. S2D). Similar junctions can be seen between all glial cells (supplementary material Fig. S2E,F).

During embryonic development the wrapping glia does not engulf individual axons (Stork et al., 2008). However, in third-instar larvae all axons are either wrapped individually or in smaller clusters (Fig. 2; Stork et al., 2008). To better dissect the process of axon wrapping, we performed a detailed temporal analysis concentrating on A1-A7 nerves (Fig. 2). These studies demonstrate that the differentiation of the wrapping glia begins at early larval stage and proceeds until puparium formation, whereas the number of axonal profiles appears to be constant during larval stages (Fig. 2). Within all abdominal nerves, different axon sizes can be recognized (Fig. 2A-F), but preferential wrapping of larger caliber axons, as described for vertebrates, cannot be noted (Nave and Trapp, 2008). We also failed to detect statistically significant distributions of stereotyped axon fascicle sizes (n=49, data not shown). To discriminate whether motoaxons are wrapped differently than sensory axons, we expressed a UAS-CD2HRP construct specifically in motoneurons employing the OK371-Gal4 driver (Dubois et al., 2001; Mahr and Aberle, 2006; Watts et al., 2004). Following a 3,3′-diaminobenzidine (DAB) staining we detected the precipitate by its strong contrast in the TEM (Fig. 2G). This analysis revealed 29-32 motoaxons, which corresponds well to the number of motoneurons determined in previous reports (n=4) (Landgraf and Thor, 2006; Sink and Whitington, 1991). Motoaxons run preferentially in common fascicles, but we did not detect prominent differences in axon size distributions compared with sensory neurons. In summary, these findings indicate that wrapping is a progressive process.

To test the molecular determinants controlling glial differentiation, we followed a candidate approach. We first tested the role of the EGF receptor (EGFR) in regulating glial wrapping in Drosophila. To circumvent the early requirements of EGF receptor signaling (e.g. during the formation of the blood-brain barrier), we first utilized the temperature-sensitive mutant egfr^tsla (Kumar et al., 1998). However, when we placed egfr^tsla mutants at the restrictive temperature in the first or second larval instar stage, no third-instar larvae were obtained. We then expressed a dominant negative form of the EGF receptor in all glial cells using the repo-Gal4 driver, and restricted the expression time to larval stages using the TARGET system (McGuire et al., 2003). The inhibition of EGF receptor signaling from first instar larval stage onwards resulted in a significantly reduced wrapping index (Table 1), and many axons were wrapped in abnormally large fascicles (Fig. 3A,B; Fig. 4). Thus, the EGF receptor needs to be activated during early larval development to allow normal glial wrapping.

We then tested whether activation of the EGF receptor is required within the wrapping glial cells or whether the EGF receptor acts in a non-cell-autonomous way to control the differentiation of the wrapping glia. Therefore, we reduced EGF receptor activation specifically in the wrapping glia by expressing UAS-EGFR^DN using the nrv2-Gal4 driver. This regime also resulted in a reduction of the wrapping index (Fig. 3A,C and Fig. 4; Table 1) but did not alter the number of wrapping glial cells along the nerve (n=5 animals, counts of nrv2≫lamGFP≫EGFR^DN-positive nuclei in 10 nerves per animal). By contrast, the expression of a dominant negative EGF receptor specifically in the SPG did not cause any abnormal wrapping phenotypes (Table 1). Similar results were obtained following expression of EGF receptor^dsRNA (Fig. 3D).

The findings described above suggest a direct relationship between EGF receptor activation and wrapping glial membrane extension. To further validate these results, we performed gain-of-function experiments. Whereas the panglial expression of activated EGF receptor (λtop) results in the induction of cell proliferation in peripheral nerves (Read et al., 2009; Witte et al., 2009), the expression of activated EGF receptor only in the wrapping glia does not alter wrapping glial cell number (n=5 animals, counts of nrv2≫lamGFP≫EGFR^λtop-positive nuclei in ten nerves per animal). In addition, we determined the ultrastructure and analyzed nine nerves of four animals. The wrapping index of these nerves was grossly normal (Table 1). In two specimens, however, we noted a dramatic increase of glial membrane production, which was never observed in control animals (Fig. 3E,E′). In conclusion, the EGF receptor is autonomously required in wrapping glial cells to instruct glial wrapping around axons.

The Drosophila EGF receptor can be activated by four different ligands: Spitz, Keren, Gurken and Vein. As spitz null mutants die during early embryogenesis, we suppressed the spitz gene function by RNA interference (RNAi) in either neuronal or glial cell types. Suppression of spitz function in neurons (using the elav-Gal4 driver) resulted in a normal wrapping phenotype (Table 1). Moreover, neuronal expression of secreted Spitz, which is known to be a potent activator of the EGF receptor, also does not influence glial wrapping (Table 1). Suppression of spitz function in glial cells (using the repo-Gal4 driver) also does not affect glial differentiation (Table 1). However, when secreted Spitz is expressed by glial cells, axonal wrapping is promoted (Table 1).

As Keren and gurken mutants are viable and do not show behavioral deficits (Brown et al., 2007; Reich and Shilo, 2002; Schüpbach, 1987), we then focused our analysis on vein mutants. vein encodes the fly homolog of neuregulin 1, which regulates the ensheathment of peripheral axons in mammals (Michailov et al., 2004). The allelic combination vein^RY/vein^L6 represents the null situation and results in lethality during early pupal stages (Donaldson et al., 2004; Schnepp et al., 1996; Yarnitzky et al., 1997).

In vein^RY/vein^L6 mutant third-instar larvae, we noted a prominent reduction of the wrapping index (Fig. 3A,F; Fig. 4; Table 1). Overexpression of vein, which is known as a weak activating EGF receptor ligand (Schnepp et al., 1996), in either neurons or glial cells did not significantly change the wrapping index (Table 1). To test whether Vein is provided by neurons or glial cells we performed cell type-specific RNAi-mediated gene knockdown experiments (Table 1). Suppression of vein function specifically in neurons did not cause any abnormal phenotype in peripheral nerves. By contrast, when we silenced vein specifically in all glial cells we observed a reduced wrapping index, similar to vein loss of function mutants (Fig. 3F,G; Table 1). This suggests that the EGF receptor is not activated by a neuronally derived signal but rather by a glia-derived signal. To identify the relevant glial cell type, we suppressed vein function either in the wrapping glia or in the SPG, which directly abuts the wrapping glia. Surprisingly, silencing of vein specifically in wrapping glial cells strongly affected axonal wrapping, indicating that an autocrine signaling mechanism is orchestrating wrapping glial differentiation (Fig. 3A,H; Fig. 4).

To further verify that Vein needs to be expressed by the wrapping glia in order to induce the axonal wrapping, we initiated cell type-specific rescue experiments. However, the basal activity of the UAS-vein construct is so high that it rescues the lethality of vein^dddL6/vein^dddRY transheterozygous flies, which precludes further genetic analysis.

The loss-of-function experiments mentioned above suggest that the EGF receptor is expressed by wrapping glial cells. Anti-EGF receptor antibodies had been generated but did not allow a faithful localization of the protein in larval tissues. Instead, we utilized the recently developed MiMIC technology, which provides an elegant approach to generate endogenously tagged protein variants (Venken et al., 2011). We generated an isoform-specific Cherry-EGF receptor fusion (Fig. 5A; and see Materials and Methods), which is expressed from the endogenous locus and which revealed significant expression in peripheral glial cells (Fig. 5B). The strongest expression was noted in the surface glia but consistent expression was also noted in the wrapping glia (Fig. 5B′). To identify cells expressing the EGF receptor ligand Vein, we utilized a Gal4 enhancer trap insertion, showing weak but consistent expression in the wrapping glia and strong expression in tracheal cells (Fig. 5C). Interestingly, spitz expression, as visualized by a Gal4 enhancer trap line, also appeared prominent in wrapping glial cells (Fig. 5D). In conclusion, these findings support the notion that axonal ensheathment is controlled in an auto- or paracrine fashion by Vein secreted from wrapping glial cells.

During the experiments described here we noted that modulation of EGF receptor activity in the wrapping glial cells also influenced the development of the outer glial cell layers that constitute the blood-brain barrier. Most prominently, the number of perineurial glial cells found on every abdominal nerve is affected. In wild-type animals, ∼15-30 glial cells associate with the nerve extension region of the A2-A7 nerves (see Fig. 1 for definition of nerve extension region), and ∼100 glial cells are found in the nerve extension region of the A8 nerve (Fig. 6C).

In control animals, which carry a copy of the nrv2-Gal4 element, glial number is generally reduced. Between 10 and 20 glial cells localize to each abdominal nerve extension region (A2-A7, for A8, >60 glial cells were counted) (Fig. 6; also see von Hilchen et al., 2013). Interestingly, a linear increase in glial cell number is noted at nerves A2-A7. As the length of abdominal nerves also increases linearly, within every additional 500 µm nerve length about three extra perineurial glial cells are found. The only exception to this is the A8 nerve, which contains more than 60 perineurial glial cells (Fig. 6C).

Upon expression of a dominant negative EGF receptor specifically in wrapping glial cells, we did not note a reduction of wrapping glial cell number. However, we found a robust reduction in perineurial glial cell number, suggesting that wrapping glia-derived signals control the proliferation of the perineurial glia during larval stages. This, however, does not explain why the perineurial glial cell number is extraordinarily higher along the A8 nerve. As the A8 nerve is significantly larger in length and diameter compared with all other abdominal nerves, and harbors ∼110 instead of 78 axons, it might well be that perineurial cell number is regulated by the increase of the nerve surface.

To directly test whether the nerve surface size regulates perineurial glial cell number, we looked for alternative ways to increase the size of the nerve diameter. The serine/threonine kinase Fray is expressed by the subperineurial glial cells and is required to regulate the extracellular volume in peripheral nerves via a conserved co-transporter mechanism (Leiserson et al., 2011,, 2000). Upon glial-specific fray knockdown, peripheral nerves show prominent swellings and thus show a local increase of the nerve surface. Consistent with the idea that nerve surface takes part in the regulation of perineurial glial cell number, we always detected a significantly increased number of perineurial glial cells that are characterized by Apontic expression in the bulged nerve areas (supplementary material Fig. S3).

In addition to this, we noted effects of the wrapping glia on the development of the SPG. In wild type, the subperineurial glial cells form autocellular contacts with pronounced septate junctions: The length of this autocellular overlap (ACO), which contains septate junctions, is ∼4.7 µm (Fig. 7A,B; n=39). Interestingly, vein mutant subperineurial glial cells show a highly significant reduction in the length of the ACO (Fig. 7A-C; n=44). Similar results were obtained when we silenced vein specifically in all glial cells (Fig. 7A,B,D; n=37). To determine which glial cell type is required for this effect, we silenced vein only in the SPG or in the wrapping glia using moody-Gal4 or nrv2-Gal4, respectively. Suppression of vein in the SPG does not affect the length of the ACO, whereas reduction of vein specifically in the wrapping glial cells results in a significant reduction of the length of the ACO (Fig. 7A,B,E; n=37 and n=35). To further corroborate these findings, we modulated the activity of the EGF receptor by either expressing a dominant negative or a constitutive active version. Whereas suppression of EGF receptor activity in the SPG results in a reduction of the ACO length, as seen in vein mutants (Fig. 7A,B,F; n=30), activation of the EGF receptor specifically in the SPG results in a prominent increase of ACO length, which is also observed following panglial expression of the activated EGF receptor (Fig. 7A,B,G,H; n=33 and n=32).

Together, these data suggest that wrapping glia-derived Vein not only regulates the degree of differentiation in the wrapping glia but also directly influences the length of the septate junction containing ACO in the SPG.

In the present study, we first demonstrated that the degree of wrapping glial differentiation in peripheral nerves is controlled by the EGF receptor. Interestingly, the activating ligand Vein is not produced by axons but rather generated by the wrapping glia itself. Subsequently, Vein generated by the wrapping glia is also able to influence the differentiation of the subperineurial glia. Lastly, the wrapping glia also regulates the proliferation of the perineurial glial cells outer surface of the nerve in a presumably indirect way. In conclusion, our data show that the wrapping glial cell appears center stage, organizing the development of peripheral nerves in Drosophila.

The development of glial cells in Drosophila is initiated during embryonic stages. Although the early definition of glial fate is well understood in Drosophila (Ragone et al., 2003; Stork et al., 2012), the molecular cues responsible for the acquisition of the wrapping glial cell fate are presently unknown. The allocation of wrapping glial cell fate must be determined very early, as the number of wrapping glial cells accompanying the segmental nerves is already specified at late embryonic stages, and their number does not change during larval stages (von Hilchen et al., 2013). Only the perineurial glia is able to divide during larval development, which in part might be regulated by the surface area of the respective nerve. Abutting the perineurial glial cells are the SPG that establish the blood-brain barrier. Surprisingly, we found that the wrapping glia is crucial for the normal differentiation of the SPG. We show that these cells express relatively high levels of the EGF receptor and directly relate the level of EGF receptor activation to the extent of ACO that contains septate junctions. As the degree of septate junction formation can determine the strength of the blocking of paracellular diffusion (Schwabe et al., 2005; Stork et al., 2008), it is likely that an increase in axonal wrapping will increase the tightness of the paracellular diffusion seal between the subperineurial glial cells.

It has been already noted earlier that expression of Gliotactin is absent in egfr and vein mutants but is expressed in spitz mutants (Sepp and Auld, 2003b). Likewise, it was shown that expression of Neuroglian is dramatically reduced in egfr mutants (Sepp and Auld, 2003b). As Gliotactin is an important constituent of tricellular septate junctions and Neuroglian is a core component of general septate junctions (Genova and Fehon, 2003; Oshima and Fehon, 2011; Schulte et al., 2003), this corroborates the pivotal function of the EGF receptor in organizing septate junctions in the SPG. A possible link between EGF receptor activation and septate junction formation might either stem from a direct transcriptional activation or it could be due to changes in alternative splicing (Edenfeld et al., 2006; Oshima and Fehon, 2011). Cell type-specific splicing of neurexinIV, encoding another core component of septate junctions, is brought about by the splicing factor How, the activity of which can be regulated through phosphorylation by Raf/MAP kinase (Nir et al., 2012; Rodrigues et al., 2012).

Interestingly, septate-like junctions are also found at paranodal junctions of myelinated nerves in mammals. In the last years, several septate junction components have been molecularly identified and were found to be remarkably conserved (Faivre-Sarrailh et al., 2004; Peles et al., 1997). Quite similarly, loss of the How homolog quaking (Qk) results in glial differentiation defects in mice (Ebersole et al., 1996; Edenfeld et al., 2006; Rodrigues et al., 2012; Sidman et al., 1964).

In Drosophila, the wrapping glial cells originate from progenitor cells of the CNS, and as soon as they are specified, they start to migrate out into the periphery, in part using the same molecular machinery that also directs the navigation of axonal growth cones (Edenfeld et al., 2007; von Hilchen et al., 2008,, 2010; Sepp and Auld, 2003a; Silies and Klämbt, 2010). At the beginning of larval life, the wrapping glial cells start to grow and extend enormous processes, reaching more than 1 mm in length. Following and in parallel to this stretch growth, axonal membranes are engulfed. Throughout our analysis we found no wrapping preference for either small or large axon calibers or for sensory or motor axons. Rather, all axons appear to have the same probability to become individually wrapped.

Axonal wrapping appears generally regulated by receptor tyrosine kinase signaling. In the developing compound eye, the differentiation of retinal wrapping glial cells is triggered by a sequential activation of the FGF receptor (Franzdóttir et al., 2009). FGF receptor expression is also well documented for Manduca glial cells, in which reciprocal communication between neurons and glial cells regulates its activity (Gibson et al., 2012). Interestingly, in Manduca it was also shown that the EGF receptor is expressed in a subset of olfactory glial cells. Global inactivation of the EGF receptor using a small inhibitor molecule does not lead to defects in glial cell migration but affects sorting of olfactory receptor axons, which also express the EGF receptor (Gibson and Tolbert, 2006). Similar reports were made regarding the role of the EGF receptor in embryonic Drosophila glial cells (Sepp and Auld, 2003b).

Thus, the activation of EGF receptor signaling appears crucial for normal differentiation of the wrapping glia. Based on expression data, we assume that this activation occurs cell-autonomously; however, an indirect mechanism might also be operating. Here, EGF receptor signaling in the CNS cortex glia would have to influence motoneuron differentiation, which in turn affects differentiation of the wrapping glia in the peripheral nerves. This would also have to influence wrapping of sensory axons and therefore appears unlikely.

Surprisingly, we only found a role for the Vein ligand that is normally acting to sustain preset EGF receptor activation (Shilo, 2005). The initial activation of the EGF receptor could be initiated in a ligand-independent manner. The broadly expressed adhesion protein Neuroglian is not only a core component of septate junctions but is also known to activate receptor tyrosine kinase signaling in a ligand-independent manner (Doherty and Walsh, 1996; Donier et al., 2012; García-Alonso et al., 2000; Gibson and Tolbert, 2006; Gibson et al., 2012; Islam et al., 2004; Kristiansen et al., 2005; Nagaraj et al., 2009). Possibly, Neuroglian might thus deliver a first trigger signal to initiate glial differentiation, independent of activating ligands such as Vein or Spitz. As soon as this interaction is set during early development, subsequent glial signaling suffices to sustain the differentiation process of the wrapping glia.

In this study, we have shown that the degree of wrapping glial differentiation in the Drosophila PNS crucially depends on the activation of the EGF receptor through an autocrine activation loop involving the neuregulin-like signaling molecule Vein. This is remarkably similar to the mechanism underlying the formation of myelin in the PNS of mammals. Here, the regulation of wrapping intensity is controlled by EGF receptor activation, too (Michailov et al., 2004; Taveggia et al., 2005). However, in contrast to what we have found in Drosophila, the activating ligand neuregulin 1 is derived from the axon. Only during injury, loss of axonal contact to Schwann cells triggers expression of neuregulin 1 in denervated Schwann cells, which serves as an autocrine/paracrine signal to promote Schwann cell differentiation and remyelination (Stassart et al., 2013). Thus, this injury model might reveal an evolutionary ancient signaling way of an autocrine activation of axonal ensheathment, which is still used during Drosophila development.

All Drosophila work was conducted according to standard procedures. Flies were kept at 25°C unless indicated otherwise. The following fly strains were used: UAS-vein2.2 (kindly provided by B. Edgar, Heidelberg, Germany; map position was determined at 94E5), gliotactin-Gal4 (Auld et al., 1995), c527-Gal4 and Mz97-Gal4 (Hummel et al., 2002), nrv2-Gal4 and repo-Gal4; repo-Gal4 (Lee and Jones, 2005; Sepp and Auld, 1999). These strains allow expression of UAS-based transgenes in glial cells. OK371-Gal4 allows to express transgenes in glutamatergic motoneurons (Mahr and Aberle, 2006), repoflp (Silies et al., 2007), UAS-λtop (Queenan et al., 1997), UAS-EGFR^DN, UAS-lamGFP and UAS-CD2HRP (Bloomington Stock Center, Bl5364, Bl7376 and Bl8763, respectively).

Single-cell clones were generated using a flip-out approach with flies in the following genotype: nrv2-Gal4, UAS-flp; UAS>CD2yellow>CD8::GFP (Wong et al., 2002). The TARGET system (McGuire et al., 2003) was used to induce expression specifically in larval stages. For this, a tubulin-Gal80^ts (tub-Gal80^ts) construct (Bloomington Stock Center, Bl7017) was combined with repo-Gal4 and the corresponding UAS-effector. Animals carrying all genetic elements were kept at the permissive temperature (18°C) until the first-instar stage and were then shifted to the restrictive temperature (29°C). The MiMIC cassette in the insertion egfr^Mi02852(Bloomington Stock Center, Bl36152) was exchanged with a Cherry exon as described (Venken et al., 2011).

The following UAS-dsRNA flies were used: UAS-vein^dsRNA [Vienna Drosophila RNAi Center (VDRC) ID 50358]; UAS-spitz^dsRNA (VDRC ID 3920); UAS-egfr^dsRNA (TRIP collection Bloomington Stock Center, Bl36770, Bl36773), all obtained from the VDRC, Vienna, Austria. Tubulin–Gal4-driven, ubiquitous expression of spitz^dsRNA resulted in lethality, whereas expression of spitz^dsRNA specifically in the developing eye using GMR-Gal4 resulted in a rough eye phenotype. Other stocks used in this study were obtained from the Bloomington Stock Center (Indiana University, Bloomington, IN, USA) or the Drosophila Genome Resource Center in Kyoto, Japan.

Fixation and treatment of tissues for immunohistochemistry was performed as described (Yuva-Aydemir et al., 2011). Anti-Repo antibodies (Halter et al., 1995; Lee and Jones, 2005) were obtained from the Developmental Studies Hybridoma Bank (Iowa City, IA, USA). Other antibodies used were as follows: anti-CD2 (1:1000; Invitrogen, MR6000), anti-GFP (1:1000; Molecular Probes, A11120), anti-Apontic (1:200; Eulenberg and Schuh, 1997) and anti-HRP 649 (1:500; Jackson ImmunoResearch Laboratories, 123-495-021). Labeled specimens were analyzed using a Zeiss 710 LSM, orthogonal sections were taken using the Zen software (Zeiss). For preparations of larval filets, larvae were raised on food containing 0.05% Bromophenol Blue. Third-instar larvae empty their gut a few hours before pupariation (Goldstein et al., 1994). Larvae with empty gut were opened along the dorsal midline and pinned onto sylgard plates. Animals were fixed using Bouin's solution. For electron microscopy analyses larvae were fixed in 4% paraformaldehyde (PFA) and 0.5% glutaraldehyde at room temperature for 2 h. The larvae were then removed from the plastic support and fixed in the same solution at 4°C overnight. Following a fixation in 2% OsO₄ for 1 h on ice and 2% uranyl acetate at room temperature for 30 min, specimens were dehydrated and subsequently embedded in epon using small insect pins. Needles were removed prior to sectioning. Sections were stained with 2% uranyl acetate for 30 min and 0.4% lead citrate for 3 min at room temperature in the dark.

For visualization of HRP-expressing cells, tissues were prefixed for 60 min in 4% PFA at room temperature. Following washes with 0.1 M phosphate buffer (pH 7.2), specimens were treated with 1% NaBH₄ for 20 min. Diaminobenzidine (DAB, 0.6%), (NH₄)₂Ni(SO₄)₂ and CoCl₂ (0.025% each) and H₂O₂ (0.06%) in 0.1 M phosphate buffer for 1.5 h. Subsequently, tissues were fixed in 4% PFA overnight at 4°C and 1% OsO₄ for 1 h on ice, followed by 30 min 2% uranyl acetate at room temperature, dehydration and embedding using epon. Ultrathin sections were directly imaged with a Zeiss EM900 with a SIS Morada digital camera.

In order to quantify the number of axons, axonal profiles were identified by their characteristic cytoplasmic staining, the shape and the presence of central microtubules, and were counted using Fiji with the cell counter plug-in. All axons with at least one direct axonal neighbor were defined as one axon bundle. The number of wrapped individual axons and axon bundles in relation to the total number of axons gives the wrapping index. A wrapping index of 100% implies that every single axon of the nerve is individually wrapped. Counting of wrapped axonal profiles was performed in part double blinded. All nerves that contained less than 76 or more than 82 axons were not included in the statistical analysis, since preparation and fixation artifacts sometimes caused loss of axons or additional axon-like structures. The typical wrapping index of wild-type wandering third-instar larvae is 19%. Significance was determined using a two-tailed Student's t-test.

We thank B. Edgar and B. Z. Shilo for flies and reagents. We are grateful to B. Risse, D. Zeuschner for help throughout the project, and instructors of the EMBO course on electron microscopy and stereology for advice. General fly stocks were provided by the Bloomington Stock Center. Antibodies were provided by Developmental Studies Hybridoma Bank and Cytoskeleton, Inc. We are thankful to FlyBase for providing information and all members of the Klämbt laboratory for support throughout the project.