Flies were reared at 25°C on standard Drosophila yeast-cornmeal molasses media. The following stocks were used: OK107-Gal4, Sema-1a^k13702, Df(2l)Exel7039, PlexB^KG00878, PlexA^{EY 16548}, UAS-lacZ.btau.YES, UAS-mCD8-Gfp, UAS-mCD8-Rfp (Bloomington stock center, Bloomington, IN, USA), Sema-1a^CA7125 and PlexA^{Y D0269} (A. Spradling, Carnegie Institution for Science, Baltimore, MD, USA), 442-Gal4 (T. Préat, Ecole Supérieure de Physique et Chimie Industrielle, Paris, France), UAS-Sema-1a-RNAi, UAS-PlexB-RNAi and UAS-PlexA-RNAi (L. Luo, Stanford University, Stanford, CA, USA; The efficacy and specificity of these RNAi constructs has been previously reported (Hu et al., 2001; Sweeney et al., 2007; Yu et al., 2010), UAS-PlexB and UAS-Sema-1a (C. Goodman, University of California, Berkeley, CA, USA), UAS-Sema-1a^Δcyt1 (T. Godenschwege, University of Massachusetts, Amherst, MA, USA), UAS-Sema-1a^Δcyt2 (A. Kolodkin, Johns Hopkins University, Baltimore, MD, USA). For mosaic analysis with a repressible cell marker (MARCM) analyses, hsFLP; Sema-1a^k13702 FRT40a/FRT40A-Gal80; Tub-Gal4, UAS-Gfp-cd8 flies were heat shocked at 37°C for 1 h at the relevant developmental stage (Lee et al., 1999).

cDNA clones for PlexA (LD10519) and PlexB (RE22882) were ordered from the Berkeley Drosophila Genome Project. In situ hybridization was done as previously described (Edwards et al., 2009). Images were obtained using a light microscope (model BX61; Olympus) and CellˆD 2.6 imaging software.

Immunohistochemical labeling of Drosophila brains was done as previously described (Rollmann et al., 2008). The following antibody dilutions were used: monoclonal 1D4 antibody (anti-FasII): 1:100; 4F3 antibody (anti-Dlg1): 1:20 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA 52242, USA); anti-GFP: 1:500 (Abcam, Cambridge, UK); anti-Sema-1a: 1:5000 (A. Kolodkin, Johns Hopkins University, Baltimore, MD, USA). FITC- or Cy3-labeled anti-mouse or anti-rabbit antibodies (1:200; Jackson Immunoresearch, Westgrove, PA, USA). Confocal imaging was performed using an Olympus FV1000 microscope. Defects in lobe length and orientation were defined by visually inspecting all brains. Only obvious and unambiguous differences in length and/or orientation were considered.

Statistical analyses were performed using Graphpad Prism 6. Statistical tests were chosen based on the format of the data (continuous or ordinal) and the distribution (Gaussian).

We used the Sema-1a^CA07125 GFP protein trap line to study the expression pattern of Sema-1a in the MBs. This line contains a P{PTT-GA} insertion between the first two exons of the gene, generating a GFP::Sema-1a fusion protein (Buszczak et al., 2007). We observe prominent expression in cell bodies throughout the cortex of the larval and adult brain as well in as axonal tracts such as the larval MBs and the adult ellipsoid body (EB) and MBs. We detected labeling in all the MB lobes of third instar larvae as well as during all pupal stages (Figure 1). We confirmed this MB expression using an antibody targeting the Sema-1a protein (Supplementary Figures S1A–C). Expression of Sema-1a in the MBs persisted throughout adulthood (Supplementary Figure S1D).

We showed that Sema-1a is expressed in the MBs during different developmental stages. To investigate the role of Sema-1a in MB development we made use of loss- and gain-of-function experiments. For these analyses, we focused on the α, β, and γ lobes as these are easy to visualize and showed the most prominent phenotypes. We also documented the effects of Sema-1a on α′ and β′ lobe development (Supplementary Tables S1–S3).

First, we made use of three mutant lines; a deficiency covering the Sema-1a locus and two P-element insertion lines, the null allele Sema-1a^k13702 and the hypomorphic Sema-1a^CA07125 allele (Yu et al., 1998; Buszczak et al., 2007). Sema-1a^k13702 contains a PlacW insertion located in the 5′ UTR of the Sema-1a gene (Yu et al., 1998). Sema-1a^CA07125 was described in Section “Sema-1a is Expressed in the Mushroom Bodies during Development as Well as in Adults.” These lines were all backcrossed into a Canton-S background to eliminate possible confounding effects due to genetic background. In the wild type Canton-S flies themselves, we did not observe any MB defects (n = 40; Figure 2A). Homozygous Sema-1a^k13702 flies do not survive beyond the first larval stage. Homozygous Sema-1a^CA07125 mutants are semi-lethal, with survivors showing strong MB defects (Figure 2E). The same range of phenotypes was seen in trans-heterozygous Sema-1a^k13702/Sema-1a^CA07125 and Sema-1a^k13702/Df(2l)Exel7039 flies (Figures 2A–C). We observed two types of defects: alterations in lobe length, resulting in short or overextended lobes and alterations in lobe orientation, resulting in ventral-medial or dorsal-lateral misoriented lobes (Figure 2A).

For further analyses, we first focused on the alterations in lobe length. Loss of Sema-1a in the investigated mutants affected the length of the MB lobes in a lobe-specific manner. The axons of the α lobes were often short, while their axonal sister branches in the β lobes displayed overextension phenotypes (Table 1; Figure 2). The γ lobes were also short. To examine the MB intrinsic requirement for Sema-1a, we looked at the effects of Sema-1a RNAi-mediated gene knock-down and overexpression using the OK107-Gal4 MB-driver (Sweeney et al., 2007; Yu et al., 2010). OK107-Gal4 drives expression in all MB neuroblasts, ganglion mother cells and neurons from embryonic stages onward (Adachi et al., 2003). Knock-down of Sema-1a in the MB neurons resulted in short α and γ lobes, reminiscent of the phenotype observed in the mutants. In contrast with the mutant lines, the β lobe phenotype shifted from overextension to shorter lobes (Table 1; Figure 3A). Overexpression of Sema-1a resulted in α lobe overextension without affecting the sister β branches (Table 1; Figures 3B,C). The γ lobes appeared to overextend. These lobes had an irregular “accordion-like” anatomy, with axon tips that grew more ventrally instead of stopping at the midline. Both α and γ lobe phenotypes are opposite to the shorter lobes observed upon loss of Sema-1a. The β lobes showed overextension in 9% of the analyzed brains. However, it should be noted that in almost all the other brains analyzed, this phenotype could not be detected due to the complete absence of β lobes, as a result of the orientation phenotype (Table 2).

These data demonstrate that the effects of Sema-1a on α and γ lobe length are MB intrinsic with loss and gain of Sema-1a leading to shorter and longer α and γ lobes, respectively. The effects of Sema-1a on β lobe length, however, seem to be, at least partially, MB extrinsic. Analyzing homozygous Sema-1a^k13702/Sema-1a^k13702 clones using MARCM partially confirmed these results, but also emphasized the complexity of the role of Sema-1a in MB development. While GFP labeled α and β lobe clones show length defects, surrounding unlabeled axons project normally (Supplementary Figures S2A–C). The short β lobe phenotype of the clones is consistent with the MB intrinsic effect observed upon RNAi knock-down. The fact that only the Sema-1a mutant clones show a phenotype in the α lobe confirms the MB intrinsic role of Sema-1a. However, contrary to the mutants and the RNAi knock-down where shorter lobes were observed, the MARCM mutant clones show overextension. We propose that these observations may reflect the need of Sema-1a levels to be carefully tuned to facilitate normal MB lobe formation or the presence of a Sema-1a modulated interaction between axons of the same lobe, or both. Previously, it has been reported that the TIFR can be involved in the regulation of midline crossing of horizontal MB lobes (Simon et al., 1998). The TIFR is a glial structure, located between the β lobes during late third instar and early pupal stages, when the β lobes are formed. Hence, we hypothesized that Sema-1a signaling in the TIFR could provide a ‘stop’ signal to growing β lobes. First, we checked whether Sema-1a is expressed in the TIFR. We labeled the TIFR by overexpressing uas-mCD8-Rfp using the TIFR-specific 422-Gal4 driver. Using the protein trap Sema-1a^CA07125, we show that Sema-1a is located in the glia of the TIFR during early pupal stages (Figure 3E). To determine whether the midline crossing phenotype of the β lobes could be due to a role of Sema-1a in neuronal- glial communication during metamorphosis, we knocked down Sema-1a in glia using repo-Gal4. Pan-glial knock-down resulted in β lobe overextension and midline crossing (16 out of 22 hemispheres; p < 0.00001). Next, we analyzed whether this effect involves the glial TIFR cells. We made use of 442-Gal4 to induce UAS-Sema-1a-RNAi and UAS-Sema-1a in the TIFR cells (Table 3; Figure 3F). Expression of both transgenes resulted in 100% penetrant β lobe overextension and fusion, while the γ lobes remained normal. This observation argues for a function of Sema-1a in the interhemispheric ring that is required for the correct development of the β lobes. Alternatively, TIFR-expressed Sema-1a may indirectly regulate β lobe formation by contributing to the development of the TIFR, whose wild-type structure might be crucial for inhibiting β lobe overextension. Analysis of the TIFR, however, showed no obvious morphological defects, indicating that Sema-1a does not have a crucial function in the development of this glial structure (Supplementary Figure S3). Furthermore, defective TIFR development has been shown to induce both β and γ lobe overextension while we observe only β defects (Simon et al., 1998). This supports a role for Sema-1a as a repulsive guidance cue in these glia rather than having a role in TIFR development. However, the fact that both knock-down and overexpression of Sema-1a in the TIFR result in β-lobe overextension suggests that Sema-1a in the TIFR may influence β-lobe development in more than one way.

Sema-1a has previously been shown to signal in a bidirectional manner, resulting in both a ligand and a receptor function for the protein (Godenschwege et al., 2002). To investigate these signaling modalities in the context of the MBs, we made use of overexpression constructs containing both the transmembrane and extracellular regions of the native protein but lacking the complete intracellular domain (Sema-1a^Δcyt1 and Sema-1a^Δcyt2; Godenschwege et al., 2002). It was previously shown that the cytosolic fragment is not required for Sema-1a surface expression, as Sema-1a^Δcyt1 is highly expressed in MB axons and is functionally active as a cell-surface ligand in the giant fiber system (Godenschwege et al., 2002; Komiyama et al., 2007). The α and γ lobe overextension phenotypes, induced with overexpression of full-length Sema-1a, were completely absent in Sema-1a^Δcyt1-overexpressing MBs (Table 1). Instead, the α lobes had a wild type morphology, while most of the γ axons were shorter than in wild type controls (Figure 3D). Similar results were obtained with a second independently generated line, Sema-1a^Δcyt2 (Table 1). These data suggest that the cytoplasmic portion of Sema-1a could be required to provide MB axons with an outgrowth signal, and thus that Sema-1a acts as an outside-in receptor in lobe outgrowth. Alternatively, the cytoplasmic domain of the Sema-1a protein could also play a regulatory role for Sema-1a as a ligand, such as fine-tuning of ligand levels or regulation of subcellular and compartment localization. Next we overexpressed Sema-1a^Δcyt1 in the glia of the TIFR using 442-Gal4. While overexpression of full length Sema-1a led to β lobe overextension, this phenotype is absent upon overexpression of the truncated protein (Table 3). These data indicate that the cytoplasmic portion of Sema-1a in the TIFR glia is required to regulate MB β lobe axon extension, and that Sema-1a could also act as a receptor in addition to acting as a ligand as suggested by the β-lobe overextension phenotype seen upon knock-down of Sema-1a in the TIFR. A receptor function of Sema-1a in glia would require additional pathways or downstream mechanisms which signal back to the β lobe axons to regulate their outgrowth. Alternatively, as suggested above, the cytoplasmic domain of the Sema-1a protein could play a regulatory role for Sema-1a as a ligand. In summary, we show that Sema-1a affects lobe outgrowth and has axonal sister branch-specific effects. Although, α and β lobes derive from the same axon, these neurites respond differentially to alterations in Sema-1a signaling. Furthermore, while the effects on the other MB lobes appear to be MB intrinsic, β lobe outgrowth appears regulated by both MB intrinsic and extrinsic Sema-1a signaling. Our data suggests a role for Sema-1a in the glia of the TIFR in this process. Finally, we show that the cytoplasmic portion of Sema-1a modulates α and γ lobe overextension.

Sema-1a^CA07125, Sema-1a^k13702/Sema-1a^CA07125 and Sema-1a^k13702/Df(2l)Exel7039 trans-heterozygotes also show a second range of phenotypes. Loss of Sema-1a results in misorientation of the MB lobes (Table 2; Figure 2). As observed with the effects of Sema-1a on lobe length, the effects on lobe orientation are also sister-branch specific. The misorientation defects observed comprised the α lobes growing more dorso-laterally than normal, or more ventro-medially resulting in growth parallel to the β lobes (Figures 2A,B). The β lobes on the other hand grew more ventrally than in wild type control brains (Figure 2A). The γ lobes showed dorso-lateral misorientation in a limited number of brains (Table 2).

To determine whether Sema-1a is required MB intrinsically for lobe orientation, we looked at the effects of Sema-1a RNAi-mediated knock-down using the OK107-Gal4 MB-driver. Knock-down of Sema-1a in the MBs had no effect on lobe orientation (Table 2). Hence, we conclude that the effects of Sema-1a on MB orientation are MB extrinsic. This was confirmed by analyzing homozygous Sema-1a^k13702/Sema-1a^k13702 clones using MARCM analyses. Both GFP labeled clones and unlabeled axons misorient (Supplementary Figure S2D). Although, MB intrinsic Sema-1a does not seem to be required for lobe orientation during normal development, overexpression of Sema-1a can influence β lobe orientation (Table 2). This effect seems to involve the Sema-1a cytoplasmic domain which might suggest a receptor function of the protein. This shows the requirement of complex fine tuning of Sema-1a in different cell types during development.

In summary, Sema-1a affects MB lobe length and orientation in a lobe and lobe sister branch-specific manner. While lobe length involves both MB intrinsic and extrinsic Sema-1a signaling, our data suggest that lobe orientation mainly depends on MB lobe extrinsic Sema-1a signaling. However, further experiments will be required to elucidate the exact mechanisms by which Sema-1a acts and whether it acts as a ligand or receptor involving only other MB axons or also other neuronal or non-neuronal cells. Due to the complexity of the observed phenotypes, the requirement of subtle fine tuning of this pathway and the broad expression of Sema-1a in the brain, correct MB development most likely relies on the integration of the combined effects of Sema-1a and the various possible mechanisms and interactions that seem involved.

We showed a complex role for Sema-1a in both MB lobe outgrowth and orientation. Plexins are well-characterized Semaphorin receptors, involved in both forward and reverse Semaphorin signaling (Whitford and Ghosh, 2001; Yu et al., 2010). In Drosophila, PlexA has been shown to function as a Sema-1a and Sema-1b receptor, while both PlexA and PlexB have been shown to genetically interact with the secreted Sema-2a protein (Winberg et al., 1998; Ayoob et al., 2006; Bates and Whitington, 2007). PlexA has been indirectly implicated in αβ lobe development through interactions with Highwire and off-track (Shimizu et al., 2011; Shin and DiAntonio, 2011). Very little is known about the expression of both Plexins in the Drosophila nervous system. By means of in situ hybridization, we show PlexA and PlexB expression in the embryonic central nervous system in concordance with previously described data (Terman et al., 2002; Ayoob et al., 2006; Supplementary Figures S4A,B). Furthermore, using the PlexA^{Y D0269} GFP reporter trap line, we show PlexA expression in the adult MB (Supplementary Figure S4C). Next, we took a closer look at the possible roles of PlexA and PlexB in MB formation.

We made use of two mutant alleles (PlexA^{EY 16548} and PlexB^KG00878), as well as RNAi constructs targeting each gene (Hu et al., 2001; Sweeney et al., 2007; Yu et al., 2010). As previously reported, heterozygous PlexA^{EY 16548} flies (homozygous lethal) showed no obvious MB abnormalities (Table 4; Shin and DiAntonio, 2011). However, RNAi-mediated knock-down of this gene using OK107-Gal4 led to strong defects in α, β and γ lobe outgrowth (Table 4; Figure 4A). In some cases, β lobe misguidance was observed (Table 4). PlexB^KG0087 homozygotes are semi-lethal and survivors had short α, β, and γ lobes (Figure 4B). Furthermore, these lobes also showed misorientation phenotypes. Some α lobes grew parallel to the β lobes, while some of these lobes grew more ventrally. Misoriented γ lobes grew more dorsally (Table 4). These phenotypes are remarkably similar to the ones seen in Sema-1a loss of function mutants. OK107-Gal4-driven RNAi knock-down of PlexB caused no neuroanatomical MB defects suggesting that PlexB functions in cells extrinsic to MBs (Table 4; Figure 4C). Our data mainly implicate a role for PlexA in stimulating outgrowth in the MB neurons, while PlexB seems to be involved in both outgrowth and orientation.

PlexA is known to be a Sema-1a ligand and has been shown to modulate reverse Sema-1a signaling in photoreceptor axon guidance (Yu et al., 2010). Hence, we wanted to investigate which effects of Sema-1a signaling in the MBs depend on PlexA. PlexB has not been implicated in MB development before, nor has it been shown to interact with Sema-1a. Introduction of PlexA^{EY 16548} or PlexB^KG00878 alleles in Sema-1a^k13702/Sema-1a^CA07125 trans-heterozygotes resulted in a profound phenotypic shift (Table 5). Both alleles caused a reduction or loss of the α lobe outgrowth defect (p < 0.0001), while β misorientation defects remained unaffected. We observed no effects on γ lobe morphology. PlexB^KG00878 introduction also led to a significant increase in ventral medial α lobe misguidance (p = 0.0385). These data confirm our findings that PlexA is mainly involved in lobe outgrowth while PlexB influences both outgrowth and orientation. We confirm the genetic interaction between PlexA and PlexB and Sema-1a combining either a PlexA^{EY 16548} or PlexB^KG00878 allele (or an RNAi construct against one of these receptors) with OK107-Gal4-driven Sema-1a overexpression. Introduction of these alleles led to a robust decrease in α overextension (Table 6). However, the correctly guided β axons failed to stop at the midline. This finding is consistent with our previous results showing a MB extrinsic effect of Sema-1a in the TIFR glia. We previously showed that MB intrinsic Sema-1a signaling is not required during development for correct lobe orientation, but overexpressed Sema-1a can influence this phenotype. We show that both PlexA and PlexB influence the effects of Sema-1a overexpression on lobe orientation, although the latter does not seem to be required for correct lobe orientation during development (Supplementary Table S4). These findings suggest a complex interaction between PlexA and PlexB with Sema-1a, which modulates different sister-branch specific effects in the different MB lobes. The complexity of this interaction is further emphasized by the suppression of the effect on α lobe outgrowth upon loss and overexpression of Sema-1a. Hence, the role of Sema-1a in the development of the MB most likely depends on careful fine tuning of MB intrinsic and extrinsic effects involving different MB axons and cell types, modulation by different receptors and interactions with other developmental pathways. Finally, we checked whether PlexA or PlexB were also involved in the regulation of β lobe overextension in the TIFR glia. Knock-down of both Plexins in this structure using 442-Gal4 had no effect on MB lobe morphology (Table 3).