Neurofibromatosis-like phenotype in Drosophila caused by lack of glucosylceramide extension

Abstract

Glycosphingolipids (GSLs) are of fundamental importance in the nervous system. However, the molecular details associated with GSL function are largely unknown, in part because of the complexity of GSL biosynthesis in vertebrates. In Drosophila, only one major GSL biosynthetic pathway exists, controlled by the glycosyltransferase Egghead (Egh). Here we discovered that loss of Egh causes overgrowth of peripheral nerves and attraction of immune cells to the nerves. This phenotype is reminiscent of the human disorder neurofibromatosis type 1, which is characterized by disfiguring nerve sheath tumors with mast cell infiltration, increased cancer risk, and learning deficits. Neurofibromatosis type 1 is due to a reduction of the tumor suppressor neurofibromin, a negative regulator of the small GTPase Ras. Enhanced Ras signaling promotes glial growth through activation of phosphatidylinositol 3-kinase (PI3K) and its downstream kinase Akt. We find that overgrowth of peripheral nerves in egh mutants is suppressed by down-regulation of the PI3K signaling pathway by expression of either dominant-negative PI3K, the tumor suppressor PTEN, or the transcription factor FOXO in the subperineurial glia. These results show that loss of the glycosyltransferase Egh affects membrane signaling and activation of PI3K signaling in glia of the peripheral nervous system, and suggest that glycosyltransferases may suppress proliferation.

Glycosphingolipids (GSLs) are essential membrane components primarily present in the outer membrane leaflet of all eukaryotic cells and are required in the development of arthropods and vertebrates (1–4). GSLs are involved in the assembly of signaling molecules, modulation of cell adhesion, protein trafficking, and differentiation (5). Targeted mutagenesis of specific murine glycosyltransferases has illustrated the fundamental importance of GSLs in the nervous system (3, 6), and mutations affecting GSL synthesis and degradation trigger severe neuropathologies in humans (7). However, the complex nature of vertebrate GSL structures and distribution patterns derived from multiple competing biosynthetic pathways makes them complicated to study, and specific molecular mechanisms for the biological functions of distinct GSL species have been difficult to unravel.

Biosynthesis of GSLs takes place in the endoplasmic reticulum and Golgi, where a variety of glycosyltransferases add monosaccharides to a growing oligosaccharide chain linked to ceramide. Vertebrate GSLs are built on either Galβ1-ceramide or Glcβ1-ceramide. Galβ1-ceramide is modified into sulfatides, whereas Glcβ1-ceramide is modified into Galβ1–4Glcβ1-ceramide (LacCer) that diverge at the third biosynthetic step to form different classes of structures differentially expressed during development and differentiation (see Fig. 4A). In contrast, one major GSL biosynthetic pathway exists in Drosophila (Fig. 1). We previously identified Egghead (Egh) and Brainiac (Brn) as glycosyltransferases responsible for GSL biosynthesis in Drosophila, catalyzing the addition of the second and third glycosyl residues of the GSL core oligosaccharide chain, respectively (Fig. 1A) (1, 2, 8). We also showed that egh and brn mutants are devoid of elongated GSLs, demonstrating that these genes are key regulators of GSL biosynthesis (2). Drosophila is therefore a very useful model system for studies of specific functions of distinct GSL species in vivo. egh and brn mutants exhibit identical phenotypes during embryogenesis and oogenesis that are reminiscent of loss of function in the Notch receptor and EGFR pathways (9, 10). Only a few studies have analyzed neurological phenotypes in egh and brn mutants, although it has been shown that egh is needed in apterous-expressing neurons for correct neural development and behavior (11).

Fig. 4.

Hypertrophic nerves with attached immune cells in is due to lack of GlcCer extension. (A) Biosynthesis of GlcCer-related glycosphingolipids in vertebrates. (B and C) Bright-field photomicrographs of larval peripheral nerves near the VNC. Brackets indicate the diameter of the right A9 nerve in (B) brn^1.6P6 and (C) after introduction of mammalian β4GalT6 (egh/Y; Gli-Gal4/UAS-β4GalT6). Note rescue of the egh phenotype in C. [Scale bar, 50 μm (B and C).] (D) Scatter plot of the A9 nerve diameter. Mean ± SEM is indicated; ***P < 0.0001. n.s., not significant. (E) Proportions of larvae with plasmatocyte accumulation on A9 nerves. Error bars mark 95% confidence intervals. The egh^62d18 allele was used in C–E.

Fig. 1.

egh mutants exhibit enlargement of peripheral nerves with attached immune cells. (A) Biosynthesis of glycosphingolipids in Drosophila. (B, C, and E) Bright-field photomicrographs of the nervous system in dissected third instar larvae. Dorsal view; anterior is upward. The bilateral nerves innervating abdominal segment 9 (A9) including their exit from the ventral nerve cord (VNC) are in focus. Other nerves originating from more anterior segments are also visible. Brackets indicate the diameter of the right A9 nerve in wild type (B) and egh^62d18 (C). Note in C the ∼50% increase in nerve diameter, and the plasmatocytes (PC) accumulating on the nerves. (D and D′) Expression of the plasmatocyte-specific marker eater-GFP in the egh background (egh^62d18/Y; eater-GFP/+) verifies the identity of the nerve-attached cells in egh as plasmatocytes. (E) Rescue of the egh phenotype by ubiquitous expression of an Egh transgene (egh/Y; UAS-Egh/+; act5c-Gal4/+). [Scale bar, 50 μm (B–E).] (F) Scatter plot of the diameter of the A9 nerves, measured 48 μm from the VNC exit. Data are mean ± SEM; **P < 0.001, ***P < 0.0001. (G) The proportions of larvae with plasmatocyte accumulation on A9 nerves. Error bars mark 95% confidence intervals.

Here we exploited Drosophila to study the effect of distinct molecular species of GSL on the development of the peripheral nervous system. We find in egh mutant larvae that peripheral nerves exhibit a dramatic overgrowth and are covered with macrophage-like plasmatocytes. This phenotype resembles human neurofibromatosis type 1 (NF1; von Recklinghausen’s disease), a common autosomal dominant genetic disorder characterized by disfiguring nerve sheath tumors with infiltrated mast cells, increased cancer risk, and learning deficits (12). NF1 is due to loss-of-function mutations in the neurofibromatosis gene encoding neurofibromin, which contains a GTPase-activating domain that inactivates the small GTPase Ras (13). Loss of this negative control in NF1 enhances Ras signaling and promotes glial growth mediated by the Ras effector phosphatidylinositol 3-kinase (PI3K) and its downstream kinase Akt (14).

We find that down-regulation of the PI3K/Akt/FOXO signaling pathway suppresses the nerve overgrowth in egh mutants. This suggests that similar to loss of neurofibromin in NF1, lack of elongated GSLs leads to aberrant activation of PI3K and overgrowth of glial cells in Drosophila.

Results

egh Mutants Exhibit Enlargement of Peripheral Nerves with Attached Immune Cells.

In Drosophila, extension of glucosylceramide (GlcCer) is exclusively controlled by the mannosyltransferase Egh (Fig. 1A) (2, 8). In third instar larvae, we found that loss of egh was often associated with enlargement of the peripheral nerves (Fig. 1 C and F). This was most pronounced in the peripheral nerves that project from the most caudal segment (A9) of the ventral nerve cord. Hence, the following analysis was focused on these nerves. Two different egh loss-of-function mutants, egh^62d18 and egh^65h5, exhibited a ∼40% and ∼20% increase, respectively, in the diameter of the nerves compared with wild type (Fig. 1F and Table S1). Another striking observation was accumulation of loosely attached cells to egh mutant nerves (Fig. 1 C and G). By expression of the plasmatocyte-specific marker eater-GFP, the attached cells were identified as plasmatocytes, a macrophage-like cell type normally circulating in the hemolymph (15) (Fig. 1 D and D′). In severe cases, large plasmatocyte aggregates completely enfolded the nerves. There was no obvious correlation between nerve diameter and the degree of plasmatocyte accumulation in egh, suggesting that these traits occur in parallel, at least at the larval stage (Fig. S1). Neither nerve overgrowth nor plasmatocyte accumulation was observed in wild-type larvae (Fig. 1 B, F, and G). Moreover, using the Gal4/UAS system (16), we found that both traits were rescued efficiently by ubiquitous expression of a UAS-Egh construct controlled by the act5C-Gal4 driver (17) (Fig. 1 E–G and Table S1).

egh Nerves Are Disorganized with an Increased Number of Subperineurial Glial Cells.

To further characterize the thick nerve egh phenotype, nerve cross-sections were investigated with transmission electron microscopy (TEM) (Fig. 2 A and B and Fig. S2). Drosophila peripheral nerves have a structure similar to mammalian unmyelinated nerves (18). The axons are ensheathed by three glial layers: an inner layer of wrapping glia (analogous to mammalian Schwann cells), a middle layer of subperineurial glia (the primary constituent of the hemolymph–brain barrier that is analogous to the blood–brain barrier), and an outer layer of perineurial glia. The nerve is surrounded by an outermost layer consisting of an acellular matrix (neural lamella) deposited by the plasmatocytes (Fig. 2A) (18). Wild-type nerves presented well-organized cellular layers with axons densely packed in wrapping glia, and clearly defined subperineurial and perineurial glia (Fig. 2A, Left). In contrast, egh mutant nerves were overgrown, and displayed loosely packed axons within an enlarged glial territory (Fig. 2A, Right). The average diameter of the A9 peripheral nerves imaged by TEM was increased by ∼80% in egh mutants compared with wild type (Fig. S2). In the most severe case observed, the nerve diameter was increased by threefold. Moreover, this nerve was encapsulated by plasmatocytes caught in the active process of what we interpret as either remodeling or attacking the neural lamella (Fig. 2 B and B′). Measurement of nerve diameters revealed that the increase in nerve diameter in egh mutants was not limited to A9 nerves but involved at least the A7 and A8 nerves as well (Fig. S2).

Fig. 2.

egh mutants have disorganized peripheral nerves with an increased number of subperineurial glial cells. (A) TEM micrographs of cross-sections of larval A9 nerves from wild type (Left) and egh^62d18 (Right), at identical magnification. The egh image was obtained using MIA. Axons, wrapping glia, subperineurial glia, and perineurial glia layers are indicated on the wild-type nerve. In the egh nerve only the axons are colored, because it was not possible to discern the glial subtypes. (B) TEM MIA micrographs exemplifying the most severe egh nerve phenotype, with a nerve diameter of 16 μm, and tightly encapsulated by attached plasmatocytes (asterisks in B and B′). Higher magnification in B′ of outlined area shows plasmatocyte-mediated deposition and/or breakdown of the outermost layer of the nerve, the extracellular matrix (plasmatocytes are marked by asterisks). (C and D) Confocal light micrographs of cross-sections reconstructed from Z-stack scans of individual peripheral nerves. Glial cell nuclei were visualized by anti-Repo (red) and subperineurial glia by Gli-Gal4/UAS-GFP (green) in wild-type (C) or egh mutant (D) background. (E and F) Intensity-inverted fluorescence light micrograph of subperineurial glial cell nuclei visualized by Gli-Gal4/UAS-GFPnls. Green stars indicate individual nuclei in wild-type (E) or egh background (F). (G) Number of subperineurial glial (SPG) cell nuclei per A9 nerve, visualized as in E and F. For both wild type and egh, data were obtained from six larvae. Data are mean ± SEM; ***P < 0.0001.

In egh nerves, it was difficult to discern the glial subtypes in the EM micrographs. However, a distinction was achieved by confocal imaging of Gli>GFP nerves immunostained with the general glial cell marker Repo (Fig. 2 C and D). Gli-Gal4 drives expression in subperineurial glia (18). Combining detection of Repo and Gli>GFP allowed reliable localization of the Repo signal to perineurial glia, subperineurial glia, or wrapping glia (Fig. 2C, Left to Right). Corroborating the EM data, the monolayer of subperineurial glia was severely disrupted in egh nerves (Fig. 2D). To assess whether the increase in glial volume in egh mutant larvae was due to hypertrophy or hyperplasia, we counted the number of subperineurial glial cells in the A9 nerves in egh mutant and control larvae expressing nuclear-localized GFP (GFPnls) driven by Gli-Gal4 (Fig. 2 E and F). Consistent with the glial increase detected by TEM and the egh requirement in subperineurial glia (see below), we found an ∼10-fold increase in the number of subperineurial glial cells in egh mutant larvae relative to wild type (Fig. 2G).

As described, we verified that the egh nerve phenotype was rescued by expression of the UAS-Egh construct controlled by the ubiquitous act5C-Gal4 driver (Fig. 1 F and G). To determine where Egh is required to prevent nerve overgrowth, we evaluated UAS-Egh expression under tissue-specific drivers. Considering both nerve overgrowth and plasmatocyte accumulation, the most efficient rescue was obtained with the subperineurial glial cell driver Gli-Gal4 and the general glial driver repo-Gal4, with Gli-Gal4 being highly efficient (Fig. S3 and Table S1). Weaker rescue was obtained with the pan-neuronal driver elav-Gal4 (19) and the motor neuronal driver OK6-Gal4 (20). The plasmatocyte-specific driver eater-Gal4 (15, 21) failed to provide any rescue. These data agree with the interpretation that egh exerts a particularly important function in the subperineurial glia.

Ectopic Ras Activation Adds to the egh Mutant Nerve Phenotype.

We hypothesized that loss of Egh activity caused overproliferation of glial cells by dysregulation of NF1 leading to enhanced activation of Ras signaling, which in turn would activate PI3K and Akt and down-regulate FOXO. It has previously been shown that complete activation of Ras through Gli-Gal4–driven expression of activated Ras, Ras^V12, leads to overgrowth of the outer glial layer (22). We therefore tested whether Ras^V12 could induce a nerve phenotype similar to egh. Expression of Ras^V12 in subperineurial glia of control larvae increased the diameter of the peripheral nerves to a degree comparable to the egh nerves (Fig. 3 A and G and Table S1), although accumulated plasmatocytes were not seen. To test whether the nerve thickening observed in egh mutants was caused by activation of the Ras signaling cascade, we expressed the fully activated Ras in the egh mutant background. The egh; Gli-Gal4/UAS-Ras^V12 larvae displayed nerves significantly wider than in both Gli-Gal4/UAS-Ras^V12 and egh control larvae (Fig. 3 B and G and Table S1). As Gli-Gal4–driven expression of UAS-Ras^V12 provides maximum activation of Ras (22), this suggests that the growth signal induced by loss of egh is not relayed solely through Ras.

Fig. 3.

PI3K signaling is required for manifestation of the egh nerve phenotype. (A–F) Bright-field light microscopy images of larval nerves near the VNC. Brackets indicate the diameter of the right A9 nerve. (A) Gli-Gal4/UAS-Ras^V12; (B) egh/Y; Gli-Gal4/UAS-Ras^V12; (C) UAS-Dp110^CAAX/Y; Gli-Gal4/+; (D) egh/Y; Gli-Gal4/UAS-Dp110^D954A; (E) egh/Y; Gli-Gal4/UAS-PTEN; (F) egh/Y; Gli-Gal4/UAS-FOXO. [Scale bar, 50 μm (A–F).] (G) Scatter plot of the A9 nerve diameter 48 μm posterior to the VNC. Data are mean ± SEM; **P < 0.001, ***P < 0.0001. n.s., not significant.

PI3K Signaling Is Required for Manifestation of the egh Nerve Phenotype.

To investigate whether the egh nerve phenotype resulted from aberrant PI3K activity within subperineurial glial cells, we first overexpressed a constitutively active catalytic domain of Drosophila PI3K, Dp110^CAAX, under control of Gli-Gal4. These larvae developed very thick irregular nerves with attached plasmatocytes (Fig. 3 C and G and Table S1). This resembled the egh phenotype, although the effect appeared somewhat stronger. Next, we tested whether down-regulation of PI3K activity would prevent nerve growth in the egh mutant background. Subperineurial expression of a dominant-negative form of PI3K’s catalytic domain, Dp110^D954A, strongly suppressed the egh nerve phenotype (Fig. 3 D and G). In contrast, subperineurial expression of Dp110^D954A in the wild-type background only had a minor effect on the nerves, showing that the thickness of the wild-type nerve does not depend on PI3K activity (Fig. 3G and Table S1). Thus, PI3K activation phenocopies the egh phenotype, whereas suppression of PI3K activity prevents its manifestation.

The catalytic subunit of PI3K converts phosphatidylinositol (4,5)-bisphosphate (PIP2) to phosphatidylinositol (3,4,5)-trisphosphate (PIP3) at the cell membrane (23). In Drosophila and vertebrates, the tumor suppressor PTEN plays an essential role in antagonizing PI3K by converting PIP3 to PIP2 (24). Overexpression of PTEN in the subperineurial glia effectively normalized the nerve phenotype in egh, whereas the recruitment of plasmatocytes was suppressed (Fig. 3 E and G and Table S1). These results show that the PIP3-generating activity of PI3K is required in glial cells for manifestation of the phenotype caused by loss of egh.

Akt is a key transducer of PI3K activity and has previously been implicated in PI3K-induced overgrowth of the outer glial layer (22). To analyze whether Akt is required for transmission of the aberrant signal associated with loss of egh, we reduced the Akt gene dosage by introducing one copy of the partial loss-of-function Akt⁰⁴²²⁶ allele. Diminishing the Akt gene dose strongly suppressed the egh nerve phenotype regarding the nerve diameter (Fig. 3G). The plasmatocytes, on the other hand, appeared to remain in egh; Akt⁰⁴²²⁶/+ (Table S1). The reason for this discrepancy is unclear. From both murine and Drosophila studies it is known that the growth-promoting effect of activated PI3K is suppressed by FOXO, a transcription factor inhibited by Akt-dependent phosphorylation (25). Accordingly, we found that overexpression of FOXO in subperineurial glia also suppressed the egh nerve phenotype (Fig. 3F). Expression of FOXO or the dominant-negative Ras, Ras^N17, only had a minor effect on the nerves in the control background (Fig. 3G), showing that the thickness of the wild-type nerves neither depends on FOXO down-regulation nor on Ras activity in the subperineurial glial cells.

The egh Nerve Phenotype Is Caused by Lack of GlcCer Elongation.

The β3GlcNAc transferase Brn controls further GSL elongation by extending the Egh product MacCer (Fig. 1A). Interestingly, the egh mutant nerve phenotype was not phenocopied by brn mutant larvae (Fig. 4B). brn mutant nerves had a diameter comparable to wild-type nerves without attached plasmatocytes (Fig. 4 D and E). Moreover, the number of GFP-labeled subperineurial glial cells was comparable to wild-type nerves in the brn background (3 ± 0; n = 6). This suggests that the nerve phenotype is either caused by loss of MacCer or by accumulation of the precursor GlcCer but is not due to the lack of long GSLs. We previously demonstrated that the egh pharate adult lethality is rescued by introduction of the corresponding mammalian GSL core using the LacCer synthase β4Gal-T6 (2) (Fig. 4A). We therefore tested whether β4Gal-T6 also rescued the nerve phenotype. Indeed, expression of β4Gal-T6 driven by Gli-Gal4 rescued both egh overgrowth and plasmatocyte accumulation (Fig. 4 D and E). These data suggest that it is not the presence of MacCer per se that is critical to sustain normal function. More likely, the egh nerve phenotype is caused by accumulation of GlcCer.

Discussion

GSLs are abundant in the brain, and altered GSL expression results in neural disorders, including seizures and axon degeneration (3). We report here that the Drosophila glycosyltransferase Egh, responsible for elongating the GSL core GlcCer, is required for controlled growth of peripheral nerves. Loss of egh caused a dramatic overgrowth of the peripheral nerves; the nerves also accumulated plasmatocytes, a macrophage-like cell type. This phenotype was not found in brn mutants, which have blocked GSL biosynthesis at the MacCer level, demonstrating that elongated GSLs per se are not important for the nerve growth.

The observed egh phenotype has several similarities with the human disease NF1, in which an intricate interplay between glial cells and infiltrating mast cells causes neurofibromas (12). NF1 is due to a loss-of-function mutation in the tumor suppressor neurofibromin. Neurofibromin inactivates the small GTPase Ras (13), and the increased Ras signaling causes glial proliferation and tumor development if unchecked (12) (Fig. S4). In Drosophila, constitutive activation of Ras in subperineurial glia causes an NF1-like phenotype with glial growth mediated by the Ras effector PI3K and its downstream kinase Akt (22). Because we observed that the egh peripheral nerve phenotype was reminiscent of NF1, we hypothesized that egh is needed for the function of neurofibromin or other factors controlling Ras signaling. Although we did not find that the aberrant growth signal induced by loss of egh exclusively functions through Ras, we verified that Ras' downstream effector protein PI3K was affected in egh mutants. Expression in subperineurial glia of an active membrane-anchored PI3K catalytic domain resulted in larvae with a nerve phenotype comparable to egh, whereas the egh phenotype was suppressed by expression of the dominant-negative PI3K catalytic domain in subperineurial glia. In addition, the tumor suppressor PTEN, which antagonizes PI3K by converting PIP3 to PIP2 (24), normalized the hypertrophic nerves in egh when expressed in subperineurial glia, thus strengthening the argument that loss of egh leads to activation of PI3K in the subperineurial glia. In addition, we found that the excessive nerve growth in egh mutants was dependent on the PI3K effector Akt, as it was suppressed by heterozygosity of Akt1⁰⁴²²⁶. Finally, overexpression in the subperineurial glia of FOXO, a transcription factor inhibited by Akt-dependent phosphorylation, suppressed the egh nerve phenotype, consistent with an overactivation of the PI3K-Akt-FOXO pathway in egh mutants.

Several mechanisms could explain how lack of Egh activates the PI3K-Akt-FOXO pathway with glial overgrowth. One is that the phenotype is caused by lack of elongated GSLs, as these play important roles in cell recognition and signaling by confining signal transduction proteins to selected membrane compartments (4). In addition, elongated GSLs directly interact with membrane proteins that modulate receptor activation via modification of receptor dimerization, recycling, and galectin-mediated clustering (26–29). Importantly, GlcCer and MacCer are minor components of the GSL repertoire in extracts from Drosophila embryos and larvae because they represent biosynthetic intermediates for elongated GSLs (30). Therefore, the biosynthetic block introduced by Egh is likely to change the composition of surface GSLs markedly (2). However, it is not straightforward to interpret an actual structure–function relationship for the Egh phenotype. Thus, brn mutants do not show enlarged nerves, arguing that elongated GSLs beyond MacCer are not essential. On the other hand, the egh mutant phenotype is rescued by the vertebrate LacCer synthase, showing that the MacCer GSL per se is not critical either. An alternative explanation for the egh nerve phenotype is accumulation of GlcCer.

Alteration in GSL metabolism with elevated levels of GlcCer has been demonstrated to up-regulate cell signaling pathways, which result in hyperproliferation of several cell types (31, 32). Accordingly, increased proliferation of blood cells and risk of lymphoproliferative disorders are seen in patients with the lysosomal storage disorder Gaucher’s disease, characterized by an accumulation of glucosylceramide and glucosylsphingosine (33). Furthermore, increased concentration of GlcCer has been implicated in human polycystic kidney disease characterized by hyperproliferative growth of kidney epithelial cells (34). The mechanism by which increased cellular concentration of GlcCer affects cell growth is currently unknown, but it has been speculated that GlcCer directly affects the phosphorylation of cytoplasmic and nuclear cell-cycle proteins (35). Our results suggest that an increase in GlcCer activates the PI3K/Akt pathway either through an effect on a membrane receptor or by affecting membrane signaling platforms. This could in turn lead to phosphorylation of cell-cycle proteins and result in hyperproliferation. Accumulation of GlcCer could also affect the concentration of upstream sphingolipids such as ceramide (Fig. S4), known to affect cell growth (36). However, targeted down-regulation of the GlcCer synthase by RNA interference experiments cause apoptotic cell death in Drosophila (37), indicating that it is unlikely that the observed nerve phenotype is due to increased concentrations of ceramide or other upstream biosynthetic intermediates.

The increased number of subperineurial glial cells in egh mutants, however, may suggest that GlcCer-based glycosphingolipids play important roles for this subset of glial cells. The subperineurial cells are known to be responsible for formation of the blood–brain barrier (18). Our data suggest that lack of elongated glycosphingolipids and increased concentrations of GlcCer promote proliferation and possibly inhibit terminal differentiation. Although speculative, this might cause a defect in the formation of the blood–brain barrier, explaining the recruitment of plasmatocytes, which in turn could deposit growth factors stimulating nerve growth. Such a relationship between recruited immune cells and neural growth has been suggested for murine models of NF1 (38). Another possibility is that plasmatocytes are recruited as part of the natural process of neural growth, because they are important for deposition of the outermost basal lamina-like structure (neural lamella) on growing nerves (39). However, the lack of plasmatocytes on the enlarged nerves in larvae with constitutively active RAS argues against this interpretation, and additional studies are required to establish the link between recruited plasmatocytes and the neural growth in egh mutants.

In summary, the present study demonstrates that accumulation of the GSL precursor GlcCer influences the central signaling pathway PI3K-Akt-FOXO, leading to overgrowth of subperineurial glial cells in Drosophila. The results provide evidence for an intricate interplay between GSLs and the essential signaling pathways influencing glial growth.

Materials and Methods

Fly Strains and Antibodies.

Flies were raised on standard cornmeal-agar medium at 25 °C. All experiments were performed on wandering third instar male larvae. Oregon-R served as wild-type control. egh refers to egh^62d18 (2, 40) supplemented with egh^65h5 (10, 40) in Fig. 1 and Fig. S1. yw brn^1.6P6 (referred to as brn) was obtained as a recombinant with Oregon-R from yw brn^1.6P6 sn² (9, 41). Akt1⁰⁴²²⁶ is a lacZ enhancer trap creating an Akt1 hypomorphic allele (42). eater-GFP (*30, II) and eater-Gal4 (*217D201, II) were from T. Tokusumi (University of Notre Dame, Notre Dame, IN) (15, 21). Gli-Gal4 (18, 22), from M. Stern (Rice University, Houston, TX), drives expression in perineurial glia (43, 44). OK6-Gal4 (II) was from M. Landgraf (University of Cambridge, Cambridge, UK) (20). UAS-Egh.Myc (II) (referred to as UAS-Egh) and UAS-β4GalT6 (II) were from S. Cohen (Institute of Molecular and Cell Biology, Singapore, Singapore) (2). From the Bloomington Stock Center we obtained act5C-Gal4 (III) (FBti0012292); UAS-GFP (FBst0001522); UAS-GFPnls (FBst0004775) (45); UAS-Ras^V12 (FBst0004847) (46, 47); UAS-Ras^N17 (FBst0004845) (46); UAS-Dp110^CAAX (FBst0025908) (48); UAS-Dp110^D954A (FBst0025918) (48); UAS-PTEN (FBtp0011853) (49); UAS-FOXO (FBst0009575); repo-Gal4 (FBti0018692) (50, 51); and elav-Gal4 (FBti0072909) (19). Mouse anti-Repo (8D12; Developmental Studies Hybridoma Bank) (52, 53) and rabbit anti-GFP (A11122; Invitrogen) antibodies were used at 1:20 and 1:200 dilutions, respectively. Alexa 488 goat anti-rabbit and Alexa 546 goat anti-mouse (Invitrogen) were used at 1:500.

Microscopy and Image Analysis.

For visualization of peripheral nerves, larvae were dissected into fillet preparations in cold HL3 medium (54). The nervous system was examined live in bright field using a Leica DMLFS microscope, and photomicrographs were recorded with a Nikon DS-5M cooled CCD camera. Nerve diameters 48 μm away from the exit from the ventral nerve cord were measured on the micrographs with ImageJ software (National Institutes of Health). Plasmatocytes expressing eater-GFP and subperineurial glial cells expressing nlsGFP were recorded or counted live with the DMLFS microscope in epifluorescence mode. Gli>GFP larvae coimmunostained for GFP and Repo were imaged with a Zeiss LSM 700 confocal microscope.

Transmission electron microscopy for examination of nerve cross-sections was performed essentially as in ref. 55. Larvae were dissected in ice-cold Schneider’s tissue culture medium and fixed by multiple rapid exchanges with 2% (vol/vol) glutaraldehyde/2% (vol/vol) paraformaldehyde in phosphate buffer (pH 7.2–7.4), followed by four changes over 60 min. After a rinse in 0.15 M sodium cacodylate buffer (pH 7.2), the specimens were postfixed in 2% wt/vol OsO₄ and 0.5% (wt/vol) K₃Fe(CN)₆ in 0.12 M sodium cacodylate buffer (pH 7.2) for 2 h. They were dehydrated in a graded series of ethanol, transferred to propylene oxide, and embedded in Epon. Ultrathin sections (80-nm) stained with uranyl acetate and lead citrate were examined with a Philips CM 100 transmission electron microscope. Digital images were recorded and processed with the ITEM software package using the multiple image alignment (MIA) module.

Statistical Analysis.

For nerve-diameter comparisons, we used JMP software (SAS Institute). Generally, diameters were obtained bilaterally from both A9 nerves in the same animal. We used a univariate ANOVA model that could accommodate also occasional larvae from which only a single measurement was available. The contrasts option was used for post hoc comparisons. Confidence levels for the proportion of plasmatocyte-positive larvae were calculated as Wilson scores (56).