The Drosophila L1CAM homolog Neuroglian signals through distinct pathways to control different aspects of mushroom body axon development

The formation of neuronal networks during brain development involves fasciculation and defasciculation of axons, integration of signaling to direct growth cones, and control of axonal bifurcation and directed outgrowth of sister branches to establish precise neuronal connections involving multiple targets. It is currently not well understood how extracellular cues and adhesive interactions are integrated spatiotemporally to control neuron subtype-specific neuritogenesis. We study this problem and its mechanistic basis in Drosophila mushroom bodies (MBs).

The MBs are prominent neuropils in the central brain of Drosophila that are important for different behaviors, including olfactory learning and memory, visual context generalization and courtship conditioning (Liu et al., 1999; Mehren et al., 2004; Davis, 2005). The intrinsic MB neurons, called Kenyon cells, are positioned at the posterior-dorsal cortex, send their dendrites anteriorly into the calyx and project their axons into the peduncle, where fibers fasciculate and project anteroventrally to the heel region (see Fig. 1A). There, larval γ neurons branch into a dorsal and a medial lobe, which are reorganized during metamorphosis to give rise to a single horizontal lobe in the adult MBs. α′β′ neurons, which are formed in late third instar larval stages, and αβ neurons, which arise during puparium formation, project axons that bifurcate and give rise to horizontal β′ and β lobes and vertical α′ and α lobes (Lee et al., 1999). Thus, the establishment of MBs covers a range of axonal development processes, such as outgrowth, branching and guidance.

One family of evolutionarily conserved cell-adhesion molecules comprises the L1-type cell adhesion molecules (L1-type CAMs). Vertebrates have four L1-CAM members (L1, CHL1, Neurofascin and Nr-CAM) (Hortsch, 2000), whereas Drosophila has only one, Neuroglian (NRG) (Bieber et al., 1989). L1-deficiency in human and mouse leads to phenotypes that include corpus callosum agenesis, mental retardation, adducted thumbs, spastic paraplegia and hydrocephalus (Fransen et al., 1995; Dahme et al., 1997; Fransen et al., 1998; Demyanenko et al., 1999; Maness and Schachner, 2007). Although considerable insight has been gained in L1-type CAM function, important challenges remain. Specifically, it is not understood how L1-type CAMs act in different molecular complexes or how they integrate various signals to control neuritogenesis.

In this paper, we identify central brain deranged alleles as mutations in the Neuroglian locus and thereby reveal a role for NRG in different steps of MB development, including fasciculation, axon outgrowth, branching and guidance. Characterization of the mutant alleles shows that the phenotypes are caused by defective cell-cell adhesion and by impaired NRG-initiated signaling. By means of transgenic rescue experiments we identify an important role for the C terminus of the neuronal specific isoform of Neuroglian (NRG-180) in axon branching and sister branch guidance and growth. We show that NRG acts upstream and interacts directly with the MAGUK Polychaetoid (PYD), revealing a novel role for a MAGUK protein during axonal differentiation. In addition, we uncover a genetic network that, apart from Nrg and pyd, includes Sema-1a, trio and Rac1 and that controls distinct aspects of MB development.

All stocks were grown on standard medium. Nrg¹, Nrg² and Nrg³ have been reported previously (Bieber et al., 1989; Hall and Bieber, 1997). The following stocks were generously provided to us: ceb⁸⁴⁹ (Nrg⁸⁴⁹) and ceb⁸⁹² (Nrg⁸⁹²) (M. Heisenberg, Universität Würzburg, Germany); FRTpyd^C5, pyd^EY04259, UAS-GFP::PYD (M. Affolter, Universität Basel, Switzerland); pyd^tam (J. Schulz, K.U.Leuven, Belgium); Sema-1a^CA07125 (A. Spradling, Carnegie Institution, Baltimore, USA); trio^M304 (B. Dickson, Research Institute of Molecular Pathology, Vienna, Austria); Rac1^J11, Rac2^Δ, Mtl^Δ (L. Luo, Stanford University, Stanford, USA); and OK107-GAL4 (S. Sweeney, University of York, UK). A second site mutation on the original Nrg⁸⁹² X chromosome was removed by recombining with an FRT19A element (Y. Kang, PhD thesis, University of Houston, 2004). We refer to this allele as FRT,Nrg⁸⁹². pyd^LL04525 was obtained from the Drosophila Genetic Resource Center (Kyoto, Japan). The Nrg-RNAi-line was from the Vienna Drosophila RNAi Center (VDRC-6688). Nrg^G00305 (henceforth referred to as Nrg^GFP) (Morin et al., 2001) and other fly stocks were obtained from the Bloomington Drosophila stock center (Indiana University, USA). For analysis of MB overexpression of NRG-180 (referred to as OK107>NRG-180), flies with the genotype UAS-NRG-180#30/+; UAS-NRG-180#28b/+;; OK107/+ were used.

Drosophila brains were dissected, fixed and immunostained as described (Clements et al., 2008). The monoclonal antibodies BP104 (anti-Neuroglian-180, used at 1:10 dilution), 1D4 (anti-Fasciclin 2, 1:100) and 4F3 (anti-Discs Large, 1:20), developed by C. Goodman, and mAbDAC2-3 (anti-Dachshund, 1:20), developed by G. Rubin, were obtained from the Developmental Studies Hybridoma Bank (under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242). The polyclonal antibody against PYD (anti-Tamou, 1:500) was a gift from R. Ueda (Mitsubishi Kasei Institute of Life Sciences, Japan). For secondary antibodies, FITC-, Cy3- or Cy5-conjugated goat anti-mouse or goat anti-rabbit (Jackson ImmunoResearch Laboratories, USA) were used at a 1:200 dilution. Brains were mounted in Vectashield mounting medium (Vector Laboratories, USA), and analyzed with a Fluoview FV1000 confocal microscope (Olympus). Image processing was carried out with Adobe Photoshop.

Genomic DNA was extracted from male flies using standard protocols. Coding sequences of Nrg³, Nrg⁸⁴⁹, Nrg⁸⁹² and wild-type Nrg were PCR amplified with primer pairs situated in Nrg intronic sequence (see Table S1 in the supplementary material). PCR products were cloned into pCR-TOPO vector (Invitrogen) and sequenced (for primers, see Table S1 in the supplementary material). Upon initial identification, mutations were confirmed in independent replicate experiments.

The GAL4/UAS system (Brand and Perrimon, 1993) was used to express UAS-NRG-180, UAS-NRG-167 and UAS-NRG::GPI transgenes (Islam et al., 2004) with the OK107-GAL4 driver (Adachi et al., 2003) to rescue mutant Nrg MB phenotypes. Progeny were collected at wandering 3rd instar larval stage, and shifted to 29°C until late pupal stage. Hemizygous Nrg³ male pharate adults were dissected for antibody staining.

For MARCM analysis, Nrg¹ and Nrg-RNAi were used to disrupt NRG function. Nrg¹ was recombined with FRT19A as described previously (Xu and Rubin, 1993). The following genotype was generated to assess Nrg-function: Nrg¹, FRT19A/FRT19A, tubP-GAL80, hsFlp; UAS-CD8::GFP/+;; OK107/+ and hsFLP/+; UAS-CD8::GFP/+; FRT2A,UAS-Nrg-RNAi/FRT2A,tubP-GAL80; OK107/+. Controls were yw, FRT19A/FRT19A,tubP-GAL80, hsFlp; UAS-CD8::GFP/+; OK107/+ and hsFLP/+; UAS-CD8::GFP/+; FRT2A/FRT2A,tubP-GAL80; OK107/+. Flies were heat-shocked for 1 hour at 37°C at different developmental stages and allowed to develop to adult stage. Brains were stained with anti-Fasciclin 2 and anti-GFP to analyze the Nrg¹ clones.

The Hemolin immunoglobulin domain (D1-D4) crystal structure (Su et al., 1998) (PDB accession number 1BIH) was used to model Neuroglian Ig1-Ig4 domains. Alignments between Hemolin D2 and D3 domains and Neuroglian and Human L1-CAM Ig2 and Ig3 domains were based on the published alignment (Sun et al., 1990). Protein structure was visualized with MolMol software (Koradi et al., 1996) (available at www.mol.biol.ethz.ch/groups/wuthrich_group/software).

S2 cells were transfected with inducible Nrg-180^wt, Nrg-180^mut849 and Nrg-180^mut892constructs and cell aggregation assays were carried out as reported previously (Hortsch et al., 1995). pRmHa-3 Nrg-180^wt containing full-length Nrg-180 cDNA has been reported before (Hortsch et al., 1995). Nrg⁸⁴⁹ and Nrg⁸⁹² mutations were introduced as follows. An EcoRI-Asp718 cDNA fragment (~3.8 kb) was excised and cloned into pTOPOXL vector digested with EcoRI and Asp718. An EcoRI-AccI fragment was replaced by a PCR-generated EcoRI-AccI fragment containing the Nrg⁸⁴⁹ (C→T) or the Nrg⁸⁹² (G→A) mutations (see Table S1 in the supplementary material for primers). An EcoRI fragment of Nrg-180 was ligated back in the pTOPOXL vector to generate full-length cDNAs, which were cloned into the pRmHa vector as Asp718 fragments.

The following alleles of PDZ-domain protein encoding genes were tested for genetic interaction with Nrg: a¹, a^k11011b, a⁷, a^EM1, baz^EH171, baz⁴, baz^G0484, cno², dlg1^G0276, dlg1^G0342, dlg1^G0456, dlt⁰⁴²⁷⁶, dsh¹, dsh⁶, pyd^C5, pyd¹, pyd^tam, scrib^j7B3, scc^E8 and ptpmeg^EY12179. Double mutant flies were analyzed for MB lobe morphology. Interactions of Nrg, pyd, trio, Rac1 and Sema-1a alleles were tested by scoring MB phenotypes in double mutant flies. Statistical analysis was carried out using a two-tailed Fisher's Exact Test, comparing presence and absence of MB phenotypes of double mutants and single Nrg mutants. A two-tailed Fisher's Exact Test was also used to evaluate MB misguidance phenotypes in flies overexpressing NRG-180 and flies in which a mutation was introduced in this overexpression background.

PCR primers are listed in Table S1 in the supplementary material. Cytosolic NRG-encoding fragments (NRG-180cyt, NRG-167cyt and NRG-180cyt ΔTYV) were PCR-amplified with full-length Nrg-180 and Nrg-167 cDNAs as templates, and EcoRI/XhoI cloned in the pGEX-5x-3 vector (Amersham Pharmacia). cDNA for the PYD constructs was generated with the Gene Amplifier PCR kit (Perkin Elmer). The fragments encoding PYD-PDZ-domains PDZ1, PDZ2 and PDZ1-2 were generated by PCR on larval cDNA and cloned into the pGEM-T-vector (Promega). The fragments were isolated from the pGEM-T-vector (BamHI/NotI for PDZ1 and BamHI/EcoRI for PDZ2 and PDZ1-2) and ligated into the pGEXmyc-vector (Grootjans et al., 2000). Full-length PYD-F was created by inserting a BglII-XhoI fragment from clone LP05923 3′ into the pGEM-T-PDZ1-2 vector. This construct was used as template to PCR-amplify PDZ3 and PDZ1-2-3. These fragments were BamHI/EcoRI digested and cloned into the pGEXmyc-vector. Production and purification of the GST-fusion proteins, SDS-PAGE and ligand overlay was carried out as described previously (Grootjans et al., 2000).

In a genetic analysis of Neuroglian (Nrg), we tested for allelism with central brain deranged (ceb) (Strauss and Heisenberg, 1993), which maps to the same cytological position (7F2-F4). Nrg (Nrg¹, Nrg² and Nrg³) and ceb (ceb⁸⁴⁹ and ceb⁸⁹²) alleles failed to complement (results not shown). Subsequently, in ceb⁸⁴⁹ and ceb⁸⁹² flies, we identified mutations in the Nrg-coding sequence (see below). Therefore, we refer to ceb⁸⁴⁹ and ceb⁸⁹² as Nrg⁸⁴⁹ and Nrg⁸⁹²throughout the manuscript. To investigate the role of Nrg in central brain development, we used Nrg⁸⁴⁹, Nrg⁸⁹², the temperature-sensitive Nrg³ [a functional null mutant at the restrictive temperature of 29°C (Hall and Bieber, 1997)], the hypomorphic Nrg^GFP line, which has a GFP exon inserted before the last exon of the two Nrg isoforms (Morin et al., 2001; Yamamoto et al., 2006), Nrg¹, a homozygous lethal null allele (Bieber et al., 1989; Hall and Bieber, 1997), and a Nrg RNAi-line (see Fig. S1 in the supplementary material).

We determined central brain defects associated with mutations in Nrg using anti-Fasciclin 2 immunohistochemistry, labeling the MB α, β and γ lobes and the ellipsoid body (EB) (Fig. 1A,B), anti-Discs Large 1, to stain all brain neuropils, and OK107-GAL4-driven CD8::GFP, which labels all intrinsic MB neurons.

In third instar larvae, no Nrg-associated phenotypes were detected. By contrast, distinct phenotypes were seen in adult brains (Fig. 1). In all mutants, α′β′ axons displayed the same defects as the αβ axons (see below), while the γ lobes were thinner. As the temperature-sensitive Nrg³ mutants do not eclose when shifted to restrictive temperature (29°C) during first larval instar, brains were dissected from pharate adults.

The morphology of the αβ lobes was examined in detail. In Nrg⁸⁴⁹ homozygous females and hemizygous males, MB αβ axons mostly grow into ball-like structures around the peduncle region instead of projecting into peduncles and then dorsally and laterally into the lobes (Fig. 1C). This phenotype has been termed axon arrest or axon stalling (Ng et al., 2002). Occasionally, milder αβ lobe defects were observed, such as one lobe (dorsal or lateral) missing. This phenotype is probably due to defects in axon branching and/or axon guidance. The latter is characterized by α axons growing along the β lobe or β axons growing along the α lobe, instead of projecting to their normal target area. FRT,Nrg⁸⁹² flies show less severe phenotypes, including missing or misguided lobes as well as full-length MB lobes that do not grow straight vertically or medially (Fig. 1D). The temperature-sensitive Nrg³ mutants revealed strong MB phenotypes, with complete or partial αβ axon stalling, and milder phenotypes such as thinner or missing lobes (Fig. 1E). Nrg^GFP flies displayed thinner or missing αβ lobes, as well as abnormalities of lobe morphology, mainly owing to partial β outgrowth defects (Fig. 1F). RNAi downregulation of Nrg in all intrinsic MB cells with OK107-GAL4 (see Fig. S1 in the supplementary material) resulted in all MBs displaying an αβ axon stalling phenotype (Fig. 1G), comparable with Nrg⁸⁴⁹. To assess whether the strong phenotypes in Nrg⁸⁴⁹ hemizygotes and Nrg-RNAi expressing flies, including the thinner γ lobes, are (partially) due to a reduction in mushroom body cell number, we counted the number of dorsal posterior cortex nuclei that are immunopositive for Dachshund (DAC), a nuclear protein expressed in Kenyon cells (Kurusu et al., 2000; Martini et al., 2000; Noveen et al., 2000). We found that the number of DAC-positive cells in these mutants was not reduced compared with control flies (see Fig. S2 in the supplementary material).

When we overexpressed the neuron-specific NRG-180 isoform using OK107-GAL4, we observed defects in MB lobe morphology. These include phenotypes caused by defects in guidance (e.g. two lobes running parallel as seen in Fig. 1H) and fasciculation (see Fig. S3 in the supplementary material).

Nrg mutants also display EB phenotypes. In Nrg³ and Nrg^GFP flies, we mainly observed a ‘ventral open’ defect (Fig. 1E), while the majority of FRT,Nrg⁸⁹² and Nrg⁸⁴⁹ mutants show a ‘split’ phenotype (Fig. 1C,D).

To define the role of NRG signaling in MB axonal development, we first determined the spatiotemporal requirements for NRG. We analyzed the distribution of Neuroglian-180 protein using the BP104 antibody (for antibody specificity see Fig. S1 in the supplementary material). NRG-180 is expressed throughout the brain at all stages (Fig. 2A,B). The staining is prominent in axons (Fig. 2A-D″), but cell bodies also reveal distinct labeling (Fig. 2E-E″). In larval, pupal and adult MB neurons NRG-180 expression is detected in all lobes. Adult MBs display a high expression level in the γ and αβ lobes, while expression is lower in the axons of the α′β′ neurons (Fig. 2C-D″).

The quantitative and qualitative differences between phenotypes seen in γ axons and the later-born α′β′ and αβ neurons (see Fig. 1) may reflect a differential NRG requirement in distinct MB neuronal subtypes. We tested this in temperature shift experiments with the temperature-sensitive allele Nrg³ in the presence of OK107-GAL4 driven GFP::CD8 to label all MB neurons (Fig. 3). No MB phenotypes were observed in third instar larvae that were shifted to 29°C from first instar larval stage (Fig. 3B). Neither did we find penetrant MB defects in adult brains of flies raised at 29°C from first until late third larval instar and at 18°C for the rest of development (Fig. 3C). When raised at 29°C between late third instar larval stage and 24 hours APF (after puparium formation), Nrg³ flies exhibited strong and penetrant MB defects (Fig. 3D), resembling the phenotypes seen in the Nrg⁸⁴⁹ and OK107>Nrg-RNAi mutants (Fig. 1). No obvious phenotypes were seen in Nrg³ mutant animals when shifted to the restrictive temperature 24 hours APF (Fig. 3E).

In summary, NRG is crucial during late larval and early pupal stages for MB development.

The phenotypes associated with Nrg mutations suggest that Nrg regulates different aspects of axon development. To determine which of these processes require Nrg and to investigate whether NRG is required cell autonomously in MB neurons, we used mosaic analysis with a repressible cell marker (MARCM) (Lee and Luo, 1999).

We analyzed neuroblast clones that expressed Nrg-RNAi or that were homozygous for Nrg¹, a previously identified protein null allele (Bieber et al., 1989). Both functional disruptions resulted in similar neuroanatomical abnormalities, including MB axon misprojection, β-axon overextension and branching defects, with axons failing to bifurcate into horizontal and vertical lobes (Fig. 4B; data not shown).

We then investigated mutant Nrg¹ single cell clones in detail, to assess the cell-autonomous roles of Nrg. The most frequently observed phenotypes were defects in axon branching (Fig. 4D) and outgrowth (Fig. 4E). Because the distinction between α/β branching defects and putative outgrowth defects with a minimal α or β sister branch could not always reliably be made, they were grouped together (summarized in Fig. 4G). Defects were mainly found in mutant αβ cells induced at prepupal stages. In these cells, axon branching and outgrowth defects were observed in 23% of cases. To a lesser extent, defects were also found in clones induced at other developmental stages (Fig. 4G). In a fraction of clones, we observed Kenyon cell bodies without axonal/dendritic outgrowth. Although this phenotype was only rarely seen in wild-type clones (2.5%), it occurred at much higher frequency in Nrg¹ clones (11.5%). Occasionally, Nrg¹ mutant cells displayed other phenotypes, such as local MB lobe defasciculation (1.6%) and axonal misprojection to other brain regions (2.2%) (Fig. 4F). These phenotypes were never observed in wild-type control clones.

In summary, these data infer a time-specific and cell-autonomous component of NRG signaling in MB-development, with NRG being involved in multiple steps of axon projection, including fasciculation, outgrowth, branching as well as guidance.

To analyze the molecular mechanism of Neuroglian function, we molecularly characterized three Nrg alleles. In Nrg³, two missense mutations were uncovered at position 735 (G→A) and at position 752 (G→A) of exon 4, resulting in G313D and V319M amino acid substitutions. These mutations were characterized by Kang (Y. Kang, PhD thesis, University of Houston, 2004) and identified independently by Yamamoto et al. (Yamamoto et al., 2006). Nrg⁸⁴⁹ has a C→T transition at position 435 in exon 4, resulting in a S213L substitution, while Nrg⁸⁹² has a G→A transition at position 404 in exon 4, causing a D203N substitution. We then determined the effects of these mutations on protein structure and function. We first modeled the mutations (Fig. 5A). G313D and V319M (Nrg³), reside in the core of the Ig3 domain, and are likely to affect protein folding and consequently protein transport leading to accumulation in the cell body (see also Hall and Bieber, 1997). S213L (Nrg⁸⁴⁹) is situated on the Ig2 domain surface where it may influence protein conformation and/or protein-protein interaction. D203N (Nrg⁸⁹²) is conserved between Drosophila Neuroglian and human L1-CAM, as well as in the Hemolin D2 domain. It is also on the surface of the Ig2 domain, and the substitution might affect protein conformation and protein-protein interaction.

We directly analyzed whether the Nrg⁸⁴⁹ and Nrg⁸⁹² substitutions interfere with homophilic interaction using cell culture-based aggregation assays. NRG-180, when transfected in S2 cells, can mediate homophilic interaction (Hortsch et al., 1998a) (see Fig. 5C,F). The Nrg⁸⁴⁹ S213L mutation completely abolished cell aggregation (Fig. 5E,F). This corresponds to the effects on homophilic interaction of a point mutation of the corresponding amino acid in human L1-CAM (H201Q) in an individual with MASA syndrome (Zhao and Siu, 1996). The Nrg⁸⁹² D203N mutation has no effect on homophilic interaction, as transfected cells can form aggregates (Fig. 5D,F) not significantly different from those mediated by wild type NRG-180 (Fig. 5C,F). Nevertheless, the identification in an individual with MASA syndrome of a similar point mutation (D202Y) indicates that the conserved amino acid is functionally important (Sztriha et al., 2000).

In summary, the strong phenotypes in Nrg⁸⁴⁹ and Nrg³ appear to be caused by defective axonal NRG-homophilic interaction or a diminished transport of the mutant protein to the cell surface, respectively. The milder phenotype in Nrg⁸⁹² mutant brains is probably caused by the disruption of other extracellular domain-mediated interactions.

Our molecular analysis suggests that the observed phenotypes may be caused by defects in homophilic adhesion or other properties mediated by the extracellular or cytoplasmic parts of the Neuroglian protein. It has previously been demonstrated that ELAV-dependent alternative splicing generates two Neuroglian protein isoforms in Drosophila. NRG-167 is ubiquitously expressed and non-neuron specific, whereas NRG-180 is neuron specific (Koushika et al., 1996). Both isoforms share the same extracellular domains, transmembrane domain and part of the cytoplasmic domain containing a conserved ankyrin binding site (see Fig. 6A for schematic representation). We performed rescue experiments of Nrg³ MB phenotypes with different transgenes to determine whether both NRG isoforms are functionally equivalent.

Hemizygous Nrg³ males and hemizygous Nrg³ males also heterozygous for OK107-GAL4, UAS-Nrg-167 or UAS-Nrg-180 display the same MB phenotype. Only a fraction of αβ-axons project to the peduncle, while the majority stalls close to the Kenyon cell bodies (Fig. 6C,F).

NRG-167 expression in a Nrg³ mutant background, using OK107-GAL4, partially rescues this phenotype. In 92% of analyzed brain hemispheres, αβ-neuron peduncles were observed (Fig. 6D). However, only 21% of α and β lobes displayed a full-length lobe phenotype. Similar results were obtained in Nrg³ hemizygous flies containing 2 UAS-Nrg-167-constructs driven by OK107-GAL4 (100% with peduncle, 8% with full-length α/β lobes) (Fig. 6F).

By contrast, expression of NRG-180 almost fully rescues the mutant phenotype; 94% of analyzed brain hemispheres have peduncles and full-length αβ lobes (Fig. 6E,F). Similar results were obtained with three different UAS-Nrg-180 lines. The extracellular domain alone is insufficient as expression of NRG-GPI does not rescue the phenotype (data not shown). All these data suggest that NRG-180 is necessary in MB neurons for normal axon projection, and that its specific cytoplasmic region is required in this process.

The rescue experiments indicate that NRG-180-specific signaling is important for MB development. The NRG-specific C terminus has an additional 63 amino acids compared with NRG-167 and contains a putative PDZ-interacting domain. Therefore, we hypothesized that NRG-180 interacts with a PDZ-domain containing protein. To identify this interaction partner, we scored MB α and β lobe phenotypes in heterozygous and hemizygous Nrg^GFP flies and heterozygous Nrg⁸⁴⁹ flies that were also heterozygous for a mutation in PDZ-protein-coding genes. Only for one gene, pyd, did all the tested hypomorphic alleles (pyd^C5, pyd¹, pyd^tam) display a prominent genetic interaction with the Nrg-alleles. Moreover, in NRG-180-overexpressing MBs, misguidance phenotypes were suppressed when a pyd mutation was introduced (see Fig. 7 and see Table S2 in the supplementary material), consistent with PYD acting downstream of NRG-180. OK107-GAL4-driven overexpression of UAS-PYD::GFP (see below) in a Nrg³ or Nrg⁸⁴⁹ mutant background, however, did not result in phenotypical rescue (data not shown).

Homo- or hemizygosity for hypomorphic mutations in pyd alone also resulted in MB lobe phenotypes. A short α lobe was found in 7% of pyd^EY04259 brain hemispheres (n=14) (Fig. 8A) and in 7% of pyd^EY00825 hemispheres (n=30), while thinner α lobes occurred in 38% of pyd^BG02748 hemispheres (n=40). In pyd^LL04525 flies, 50% of α and β lobes were thin or missing (Fig. 8C), 20% of β lobes displayed an overextension phenotype, 10% of α lobes was short and in 5% of α lobes, a misguidance defect was observed (Fig. 8D) (n=20). In flies hemizygous for pyd^KG02008, pyd^C5 and pyd^EY04259, missing lobes were detected in 14% (n=22) (Fig. 8B), 5% (n=22) and 5% (n=22) of brain hemispheres, respectively. We did not observe any obvious defects in the gamma lobes.

Since the genetic interaction data imply that NRG and PYD are components of a signaling pathway active in MB development, we determined whether NRG and PYD (1) colocalize in the MBs and (2) physically interact. In the late larval and early pupal brain, PYD expression levels were high in the optic lobe proliferation centers and cells in the pars intercerebralis, and lower throughout the rest of the central brain (Fig. 8E; see Fig. S1 in the supplementary material). PYD immunostaining on flies expressing GFP::CD8 with an OK107-GAL4 driver, revealed colocalization of PYD immunoreactivity with GFP on the plasma membrane of the Kenyon cell bodies (Fig. 8F,G). In addition to PYD being detected in Kenyon cell bodies, we also obtained evidence for a putative axonal PYD localization in these cells. In neurons where PYD is expressed at higher levels, such as in cells in the pars intercerebralis, PYD can be readily detected in cell bodies, whereas lower protein levels are visible in the axons (see Fig. 8G). In addition, we analyzed the subcellular localization of GFP::PYD, a fusion protein that has a subcellular organization similar to endogenous PYD (Jung et al., 2006). The fusion protein was expressed in larvae and adults with OK107-GAL4, which drives expression in MB cells and pars intercerebralis cells. In both cell types, PYD::GFP was present in cell bodies as well as in axons (Fig. 8G′ and 8G″).

In summary, we showed that NRG and PYD colocalize in Kenyon cells during late larval and early pupal stages, time periods during which NRG is required for MB development.

Finally, to analyze physical interaction of NRG-180 and PYD, we used ligand overlay experiments with the cytosolic parts of NRG-167 (NRG-167cyt), NRG-180 (NRG-180cyt) and NRG-180, which lacks the last three amino acids of the PDZ-interacting domain (NRG-180cytΔTYV) that are predicted to be crucial for PDZ-domain interaction (Hung and Sheng, 2002), and with GST-myc-fusion proteins, which contain one or more of the three PYD-PDZ-domains. NRG-180cyt-binding with myc-tagged PDZ-2, PDZ-3 and the myc-fragment did not exceed background levels, which are caused by GST-GST dimerization (Fig. 8H). However, NRG-180cyt, but not NRG-167cyt or NRG-180cytΔTYV, exhibited an elevated binding affinity for myc-tagged PDZ-1-2-3, PDZ-1-2 and PDZ-1. In summary, the cytoplasmic domain of NRG-180 binds directly to the first PDZ domain of PYD.

The transgenic experiments show that peduncle formation depends on the part of the NRG protein shared between NRG-167 and NRG-180, as the axonal stalling phenotype is rescued by both isoforms. Normal lobe morphology, however, requires the additional 63 amino acids present in the NRG180 isoform. For the latter, we have shown that the interaction with PYD is required. The presence of axonal stalling and axon guidance phenotypes stimulated us to identify additional components of the signaling mechanism(s) that involve Nrg by means of genetic experiments and a candidate gene approach.

Axonal stalling phenotypes have previously been seen in mutants of Rac1, a member of the Rho family of small GTPases (Ng et al., 2002), and trio, a guanine nucleotide exchange factor essential for RAC1 activation (Awasaki et al., 2000; Hakeda-Suzuki et al., 2002). Moreover, the fact that many signal transduction pathways involved in axonal morphogenesis converge on Rho GTPases (Huber et al., 2003) make RAC1 and TRIO plausible downstream effectors of NRG signaling in MB axon development.

The axon guidance phenotypes seen in Nrg mutants raise the issue of whether specific axon guidance molecules may be involved. Several lines of evidence suggest that Sema-1a (Yu et al., 1998) may be a good candidate. L1-CAM, the vertebrate homolog of NRG, together with Neuropilin and Plexin, mediate axon guidance by forming a receptor complex for secreted SEMA-3A (Castellani et al., 2000; Castellani et al., 2002). In C. elegans, the NRG homolog LAD-2 has been found to be involved in Sema2 signaling (Wang et al., 2008), whereas, in Drosophila, evidence has been presented that Nrg interacts with Sema-1a during giant fiber system development (Godenschwege and Murphey, 2009).

We analyzed MB phenotypes in heterozygous and hemizygous Nrg^GFP flies and heterozygous Nrg⁸⁴⁹ and FRT,Nrg⁸⁹² flies that were also heterozygous for a mutation in either Sema-1a, Rac1 or trio. We observed genetic interaction for each of these genes with at least two Nrg alleles (see Fig. 7 and see Table S2 in the supplementary material). Introducing null mutations of two other RacGTPases, Rac2 and Mtl, in an Nrg mutant or an Nrg,Rac1 mutant background did not have any effect on the MB phenotypes (data not shown).

Misguidance phenotypes of NRG-180 overexpressing MBs were suppressed when Sema-1a or trio mutations were introduced, whereas a mutation in Rac1 resulted in exacerbated MB defects (see Fig. 7; see Table S2 in the supplementary material).

Finally, we tested for interactions among the four different Nrg interactors we identified. We detected strong genetic interaction between the pyd and Rac1 alleles, mainly characterized by axon stalling and guidance defects (see Fig. 7; see Table S2 in the supplementary material). Genetic interaction was also observed between the Sema-1a^k and the pyd^C5 alleles, with double heterozygotes displaying axonal guidance and overextension defects (see Fig. 7; see Table S2 in the supplementary material).

In the current study, we have demonstrated a requirement for Neuroglian signaling in different steps of mushroom body (MB) axonogenesis, namely (1) axonal projection into the peduncle, and (2) branching, outgrowth and guidance of axonal sister branches. The two steps in mushroom body axonogenesis are genetically separable and seem to involve distinct NRG signaling complexes.

In peduncle formation, NRG signaling does not rely on the NRG-180-specific intracellular domain, but on the extracellular domain and the part of the cytoplasmic domain common to both NRG isoforms. The extracellular domain contributes intercellular adhesive properties, necessary for axon fasciculation into a peduncle. This conclusion is supported by the defective adhesive properties of the NRG⁸⁴⁹ mutant protein in cell aggregation assays, and by the fact that Nrg⁸⁴⁹ hemizygotes frequently lack the peduncle. Interaxonal fasciculation in the peduncle probably involves binding to and stabilization by the actin cytoskeleton network via the ankyrin-binding domain shared by the two NRG isoforms. This conclusion is supported by previous aggregation experiments in Drosophila S2 cells in which it was shown that homophilic binding of NRG leads to recruitment of ankyrin to the contact sites (Dubreuil et al., 1996; Hortsch et al., 1998b) and by the observation that RNAi-mediated knockdown of neuron-specific ank2-RNA results in MB phenotypes similar to those seen in Nrg mutants (T.G. and P.C., unpublished).

MB lobe development, on the other hand, requires the NRG-180-specific intracellular fragment. We showed that PYD acts downstream of NRG-180 during the formation of α and β lobes. Consistent with this, we never observe axon stalling defects (i.e. lack of peduncle formation) in pyd mutants, whereas we do see defects in lobe outgrowth, branching and guidance. Furthermore, the neuron-specific NRG-180 isoform can bind directly to the first PDZ-domain of this MAGUK protein. Our observation that NRG and PYD interact to mediate proper sister neurite projections defines a novel role for the ZO-1 homolog PYD in axonogenesis. Thus far, the best-known role of MAGUKs in the nervous system has been in synapse development and function, as is the case for one of the prototypic MAGUKs, Drosophila Discs Large 1 (Dlg1) (Tejedor et al., 1997; Thomas et al., 1997), whereas PYD is known as a component of adherens junctions (Jung et al., 2006; Seppa et al., 2008).

Sema-1a, trio and Rac1 were also found to be a part of the genetic network that interacts with Nrg. The observation that heterozygosity for mutations in Sema-1a and trio both suppress NRG-180 overexpression induced MB phenotypes indicates that Sema-1a and trio are genetically downstream of Nrg and possibly in the same pathway. By contrast, the introduction of a Rac1 mutation in a Nrg gain- or loss-of-function background results in both cases in enhancement of MB phenotypes (see Fig. 7; see Table S2 in the supplementary material). This argues against a one-to-one signaling model between NRG and RAC1, in which RAC1 acts only downstream of NRG-180. Consistent with the genetic data, we could not detect a direct physical interaction between NRG-180 and RAC1 (data not shown), but preliminary co-immunoprecipitation data suggest that NRG-180 can bind to TRIO (Y.Y.K. and P.C., unpublished). Further experiments will be necessary to assess whether this binding also occurs in vivo, and whether it is instrumental for NRG-180-dependent modulation of RAC1 signaling.

Contrary to the observed genetic interaction between Nrg and Sema-1a, we found no evidence for interaction between Nrg and two genes that code for well-characterized Semaphorin receptors, plexin A and plexin B (data not shown). This is an unexpected observation in light of the fact that Sema-1a and plexin A and plexin B interact during mushroom body development (Liesbeth Zwarts, T.G. and P.C., unpublished). Therefore, this suggests that during mushroom body development Sema-1a acts both in a plexin-dependent and a plexin-independent way. A plexin-independent role in axon outgrowth has previously been described for vertebrate Sema7a (Pasterkamp et al., 2003).

The distinct requirement for NRG in peduncle and lobe formation is reminiscent of what has been shown for DSCAM. This protein has an early and essential role for selective fasciculation of young axons in the peduncle and is subsequently required for bifurcation and branch segregation (Wang et al., 2002; Zhan et al., 2004). In light of this, it is interesting to note that the different cell-adhesion molecules that have been implicated in MB development have a different MB expression pattern or temporal requirement for MB development. NRG is expressed in the MBs throughout its entire development, but we did not find any essential function for Neuroglian in larval MB development. By contrast, DSCAM expression disappears with fiber maturation and mutants have larval MB phenotypes (Zhan et al., 2004). Likewise, Fas2 mutants display larval lobe defects, whereas no (Cheng et al., 2001; Kurusu et al., 2002) or almost no (Fushima and Tsujimura, 2007) lobe defects were found in adult mutants. Taken together, these observations suggest that different steps in MB axonogenesis depend on combinations not only of isoforms of the same cell-adhesion molecule (e.g. DSCAM) but also of different cell surface molecules (e.g. NRG, FAS2 and SEMA-1A). Jointly, the cell-surface molecule complement of any given axon combined with guidance signals will then control the signaling required for proper neural circuit formation in the MBs.

This work is supported by FWO grant G.0285.05 to P.C., and by grants from the NICHD (RO1HD29388) and the NSF (IBN-0132819) to M.H. We thank the following people for antibodies and fly stocks: Markus Affolter, Martin Heisenberg, Allen Spradling, Barry Dickson, Liqun Luo, Ryu Ueda, Joachim Schulz, Jennifer Zallen and Sean Sweeney. We are grateful to Mimi Deprez for help with preparing the manuscript and figures, and to Glen Legge for instructions on how to use the MolMol software. We acknowledge Carlos Dotti for critical reading of the manuscript. Deposited in PMC for release after 12 months.