In total, we identified 43 GS lines causing da dendrite phenotypes at embryonic (8) or larval (35) stages. The phenotypes generated by many GS lines resemble those of known mutants and fall into five categories that may relate to known molecular pathways (Table 1). In the first category, four GS lines had phenotypes that resemble genetic mutations in shrub, which leads to reduced embryonic da dendrite growth 26. The second category (six lines) resembles phenotypes of Polycomb Group genes like E(z), esc, or Su(z)12, which are involved in the maintenance of dendritic arbors of class IV neurons 27. The third category contains seven GS lines that showed effects similar to cut mutations, the levels of which regulate class-specific dendritic growth and terminal branching 28. A fourth phenotypic category is represented by GSd484, which resembled spineless mutants because it increased class I and reduced class III and IV dendrites 29. However, spineless mutants also have increased class II dendrites, which we did not observe with GSd484. Fifth, GSd496 showed reduced dendritic branching in embryonic da neurons, similar to constitutive activation of the GTPase cdc42 15.

In addition to these five phenotypic categories, we also identified 24 GS lines causing phenotypes that did not resemble known mutants. These lines suggest the existence of additional genetic pathways that underlie dendritic development and may, via their insertion sites into the genome, provide clues about their molecular nature.

To compare dendrite morphogenesis between peripheral sensory neurons and central neurons, we executed a comparable misexpression screen for central neuron dendrites. Using a new FLPout based system we expressed the same 141 GS lines discussed above in selected RP2 motor neurons, again using mCD8::GFP to reveal dendrite morphologies 22. With this system, misexpression is initiated by 14–15 hours after egg laying (AEL). At this time, RP2 neurons have begun to establish characteristic dendritic trees in a particular neuropile territory and, in the periphery, RP2 axons have made contacts with their target muscles 30. We examined RP2 dendritic trees more than 10 hours later, at 25–31 hours AEL, when the majority of dendritic branches are normally located in the lateral neuropile and only a few branches project towards the midline (Figure 4a–c).

We identified 60 GS lines that affect specific aspects of RP2 dendrite development (summarised in Table 2; for a comparative summary of both screens see Additional file 1). The resultant RP2 phenotypes fall into two partially overlapping classes: altered dendritic growth and/or branching; and aberrant dendritic targeting.

Of the various aspects of dendritic development, directed growth into a particular territory is arguably least understood and few genes required for this process have been identified 63233343536. Dendritic targeting phenotypes recovered in this screen can be sorted into four categories: shifts of the dendritic territory to the anterior, posterior, medial or lateral. These categories are compatible with a model of dendrites being targeted along Cartesian co-ordinates. The existence of distinct dendritic domains in the antero-posterior neuropile axis has been illustrated previously 30. The phenotypes of this screen further suggest that motor neuron dendrites might also be patterned with respect to the ventral midline, along the medio-lateral neuropile axis.

Targeting relative to the ventral midline has been documented for axons 3738. Some of the molecular cues involved in this process (Slit and Netrins) and their receptors (Robo, Robo2, Robo3 and Frazzled/DCC) have also been shown to regulate midline crossing of dendrites in the Drosophila nerve cord 3334. We found that a GS insertion near lea/robo2, which encodes a receptor for the midline repulsive cue Slit 3738, caused a reduction of dendrites innervating the medial neuropile (Figure 6e). We therefore asked if midline-derived guidance cues such as Slit and Netrin might be involved in targeting dendritic trees to distinct medio-lateral territories, in addition to their documented role in midline crossing. To test this hypothesis we over-expressed other elements of the Slit and Netrin signalling pathways: robo, commissureless and frazzled. As previously reported, expression in RP2 of the robo antagonist commissureless 3940 or frazzled could lead to inappropriate midline crossing of RP2 dendrites 33. However, we found that expression of frazzled, though not commissureless (N = 49), also led to an expansion of the RP2 dendrites innervating the medial neuropile (Figure 6a–c,g), which is normally occupied by dendrites of other motor neurons such as RP1 and RP3. Conversely, expression of the Slit receptor Robo led to an absence of medially positioned branches (Figure 6f,g).

Our results support the idea that dendrites in the CNS are targeted along the antero-posterior and medio-lateral neuropile axes using, at least in part, guidance cues that also pattern axon trajectories.

We next investigated the strategy with which motor neuron dendrites innervate particular neuropile territories. Do dendritic arbors expand until their target territory is occupied? Or do motor neurons have a program of dendritic growth and branching that is independent of the positioning of dendrites within the neuropile? To distinguish between these alternatives, we altered the dendritic territory of RP2 by misexpression of an activated form of Robo (Robo-Y-F 41). This manipulation suppressed the establishment of dendritic branches in the medial neuropile anterior of the axon and led to a concomitant posterior expansion of the arbor, phenocopying GS lines that also reduce the medial dendritic territory (Figure 7). We measured the maximal distances (extent) to which dendritic trees extend anterior and posterior of the axon. We found that this manipulation led to a significant increase in the extent to which RP2 dendrites project posteriorly (dendritic extent of posterior arbors: 7.4 ± 1.5 μm wild type versus 12.7 ± 1.3 μm Robo-Y-F, p = 0.0001, t-test; Figure 7). However, expression of UAS-robo-YF does not abolish the establishment of anterior dendrites in the lateral neuropile and the anterior extent of the arbor is therefore comparable to controls (anterior arbors: 10 ± 0.9 μm wild type versus 9.4 ± 1 μm Robo-Y-F, p = 0.22, t-test; N = 8; Figure 7). This correlation between the induced absence of branches in the medial anterior neuropile and the extension of the posterior territory is compatible with the notion that expression of UAS-robo-YF may have caused a displacement of part of the dendritic tree from a medial anterior to a lateral posterior domain.

To further investigate the relationship between dendritic growth and targeting, we compared controls to RP2 neurons with marked dendritic mistargeting phenotypes as induced by misexpression of GSd421 (Figure 8). For a quantitative readout we reconstructed dendritic trees from three-dimensional confocal image stacks using recently developed reconstruction algorithms 4243 (Figure 8a,b). As anticipated, we found significant differences in the directionality of dendritic growth (that is, targeting), reflected by the maximum dendritic path lengths from the axon to the perimeter of the dendritic trees (19 ± 3 μm control versus 31 ± 4 μm experimental, p < 0.003, t-test) (Figure 8c). However, control and GSd421-misexpressing RP2 neurons did not differ significantly in other aspects of dendritic growth and branching, including total dendritic length (157 ± 16 μm wild type versus 143 ± 21 μm GSd421) and number of dendritic tips generated (35 ± 6 wild type versus 36 ± 4 GSd421).

These observations demonstrate that dendritic growth and branching are regulated by mechanisms that are genetically separable from targeting: RP2 motor neurons generate a set quantity of dendritic length and branches independent of the neuropile domains in which they are positioned. The quantitative analysis further addresses the longstanding issue of which attributes of neurons are genetically specified and which are subject to non-genetic influences and, therefore, highly variable 4445. Here we show that total dendritic length, branch point number and territories of branching are reproducible features. In contrast, the number of primary branches was highly variable.

Next, we assessed the reliability of the screening method by making use of the fact that, for each of the GS lines, the closest gene predicted to be expressed in response to Gal4 has previously been identified 23. First, in the entire collection of 141 lines, there were 29 genes for which there were at least two independent GS insertions. We determined the frequency with which independent GS-lines near the same gene gave concordant results: 20/29 genes (69%) in the da screen and 17/29 (59%) in the RP2 screen (Tables 1 and 2). Where different phenotypes were induced by the expression of independent GS lines near the same gene, this may in some instances be due to insertion site-specific variations in gene expression levels. In other cases, different GS insertions in the same gene may generate functionally distinct transcripts, as is predicted for insertions in the Kr-h1 (GSd113 = Kr-h1-RA; GSd204 = Kr-h1-RB transcript) and pnt loci (GSd229 = pnt-RC; GSd236 and GSd420 = pnt-RB transcript) 464748. With 69% and 59% concordance, respectively, and evidence for isoform-specific insertions at some of the discordant loci, we judge these screens to be a reliable means for identifying genes that influence dendrite development.

Secondly, we asked if the screens reported here had recovered predicted candidate genes. Among the 35 genes identified in the da screen, five have been studied previously in the context of da dendrite development: abrupt (ab), nanos (nos), bonus (bon), E(bx) and tramtrack (ttk). In all of these cases, the GS misexpression phenotypes are conversely related to the reported mutant or RNA interference (RNAi) knockdown phenotypes 11495051. For instance, mutation or RNAi knockdown of ab increases arborisation of class I neurons, while expression of GSd324 (inserted closest to ab) in class II-IV da neurons reduces dendritic branching as previously reported for misexpression of ab 5051. This concurrence suggests that the additional 30 candidate genes identified in the da screen may reveal new molecular determinants of PNS dendrite morphologies.

For the screen on central neuron dendrites we had to gauge its utility differently, since genes regulating the development of RP2 dendrites remain largely unknown. We therefore examined with in situ hybridization whether identified genes were actually expressed in the embryonic CNS during the time of dendritic outgrowth, stages 14–16. We assayed 47 of the 51 genes and found 43 to be expressed in the CNS, 25 of these in subsets of cells (Table 2; Additional file 3). These expression data suggest a high degree of confidence in the validity of the screen, though further loss-of-function studies are needed to test this directly.

One of our aims was to ask whether the development of peripheral and central neuron dendrites is influenced by common or distinct molecular mechanisms. We did so by comparing the effects of the same GS lines on da and RP2 neurons. Of the 35 unique genes identified in the da screen and 51 for RP2, there were 24 genes (39%) that were capable of influencing dendrites in both cell types (Table 3; Figure 9). Therefore, there were 11/35 genes (31%) that were particular to the da screen, while 27/51 genes (53%) were specific to the RP2 screen. This provides evidence that neither screen was prone to chronically low levels of gene misexpression that might prevent detection of phenotypes, and that each screen could reveal unique genes.

Finally, we wanted to test whether there were molecular mechanisms uncovered by these screens that were required for dendrite morphogenesis, and whether these mechanisms shed light on common or distinct pathways for peripheral and central dendrites. In both the da and RP2 screens, we identified several candidate genes (Kr-h1, bon, Hr38) related to signalling from nuclear hormones and particularly ecdysone (Additional file 4). In insects, ecdysone initiates major developmental transitions and regulates dendrite regression, pruning and re-growth among sensory da neurons and motor neurons, as well as central mushroom body neurons and peptidergic neurosecretory neurons 9525354555657. Kr-h1 is a stage-specific modulator of the prepupal ecdysone response 46, and ectopic Kr-h1 dramatically reduces dendrite branching in da neurons (this study, GSd113) and mushroom body neurons 58. bon codes for a transcription factor regulating genes involved in ecdysone responses 59, and Hr38 encodes an orphan receptor that can compete with the ecdysone receptor (EcR) for binding to its obligate co-receptor, Ultraspiracle (Usp) 6061.

Our findings that Kr-h1, bon and Hr38 may influence the morphologies of da dendrites suggested a new role for ecdysone signalling in addition to its role in metamorphosis when it induces dendrite regression and pruning. This is consistent with a recent study that showed that arborisations of class I da neurons are reduced by RNAi knockdown and mutations in EcR and usp 11. However, it remained unresolved whether EcR and Usp are required cell-autonomously in da neurons, as is the case for dendrite pruning at pupariation 954.

To investigate this further, we confirmed expression of the EcR-A and EcR-B1 protein isoforms 62 in dorsal da neurons of third instar larvae (Additional file 5) 54. Focusing on the class IV neuron ddaC, we then tested with three approaches whether there is a requirement for the EcR in ddaC dendrite development. First, since mosaic analysis with a repressible cell marker (MARCM) could not be applied directly to the EcR due to its cytological location, we generated MARCM clones for mutations in the EcR co-receptor, Usp 963. Relative to control clones (Figure 10a), the dendritic arbors of ddaC neurons in usp² null mutant clones had reduced numbers of branches (control = 754.1 ± 15.1 versus usp² MARCM = 580.3 ± 23.1; p < 1e^-6, t-test; Figure 10b,c). We also examined MARCM clones for usp³ (a hypomorphic allele) and usp⁵, a missense mutation in the second zinc finger of the DNA binding domain of Usp 9. Unlike the null usp² allele, neither usp³ nor usp⁵ showed branching defects in ddaC neurons (not shown), likely due to residual Usp function in these mutant clones. Second, we used the class IV-specific driver ppk1.9-GAL4 17 to express an RNAi-inducing construct of the EcR (UAS-IR-EcR) that targets all EcR isoforms 64. This resulted in a significant reduction in the mean density of branches to 75.8% of control levels, as measured by the number of branch ends normalised to dendritic field area (Figure 10e,g). Third, as an additional direct manipulation, we expressed a dominant negative form of the receptor (EcR-DN) that binds Usp normally, but fails to bind ecdysone and cannot activate transcription, and is thereby a competitive inhibitor of all endogenous isoforms of EcR 65. This also reduced the density of ddaC dendrites (50.8% of controls; Figure 10f,g), affecting primarily higher order branches and so having negligible impact on the field area (mean area in controls = 304,899 ± 7,115 μm² versus EcR-DN = 308,802 ± 7,400 μm²).

Together, the reduced branching observed in usp² MARCM clones and in the direct EcR manipulations using RNAi and EcR-DN support the idea that ecdysone promotes the arborisation of developing ddaC neurons prior to and in addition to its role in metamorphosis. This is consistent with other experiments showing that ecdysone enhances neurite outgrowth of cultured Drosophila neurons in vitro 66. We explored whether this role for the EcR could influence dendrite arborisation in embryonic and early larval development, or whether it was restricted to later larval life. We found that the EcR was expressed in embryonic dorsal da neurons, including ddaC (Additional file 5), and that reduced branching caused by EcR-DN was already apparent in first-instar larvae 28–30 hours AEL (mean branches per ddaC neuron in controls = 190.4 ± 5.0 versus EcR-DN = 139.1 ± 4.1; Figure 10h–j).

Central RP2 dendrites were also affected by GS lines near the three genes Kr-h1, bon, and Hr38 (Additional file 4). However, unlike in da neurons, misexpression of EcR-DN in the RP2 neuron using the FLPout system did not consistently affect its dendrites despite high expression levels (not shown), suggesting that EcR may not play a role in CNS dendrite growth in embryonic and early larval stages.

This study is the first published report of a misexpression approach used to identify candidate genes required for dendrite development. Clearly, this strategy has limitations since genes that induce phenotypes may not be required for dendrite development, nor are the phenotypes necessarily informative. Despite these concerns, we judge these gain-of-function screens a useful alternative to forward loss-of-function genetic screens, which can be limited by technical challenges (for example, when targeting specific cells) and by genetic redundancies. We deem it a valid strategy for gene discovery in many cases since at least 5 of the 35 genes identified in the da neuron screen (ab, nos, bon, E(bx) and ttk) are known to be required for da dendrite patterning and have loss-of-function phenotypes that are the converse of the gain-of-function phenotypes described here 11495051. Notably, the da screen led to the discovery that signaling through the EcR pathway promotes the arborisation of developing da neurons prior to metamorphosis.

Considerably more candidate genes, 51, were identified in the central neuron (RP2) dendrite screen. This difference in sensitivity between the two screens may reflect true differences in developmental robustness, or a bias within the GS collection, or technical issues such as the strength of the Gal4 drivers or the ease of detection of mutant phenotypes. The collection did not contain any genes shown previously to be required for dendrite development in central neurons (for example, shot/kak,fra,robo, sema1a, Dscam) 3233343667. We were therefore unable to validate the RP2 screen in the same manner as the da neuron screen. However, the phenotypes recovered were specific for particular aspects of dendrite growth, branching and targeting, and as such reveal the constraints of the underlying organizational framework. By way of verification we were able to show that at least 43 of 51 candidate genes from the RP2 screen are expressed in the CNS during the period of dendritic outgrowth and targeting. Since their ectopic and/or elevated expression in RP2 is sufficient to mis-pattern RP2 dendrites, these genes could encode novel factors required for CNS dendrite development.

The second aim of this study was to gain a better understanding of the organizational framework that underlies dendrite development. In both screens the misexpression phenotypes fell into specific categories, demonstrating that specific features of dendrite development, such as growth, branching and targeting can be reproducibly and selectively modulated. Within each screen, there was incomplete overlap between phenotypic categories, suggesting that these features are derived from molecularly distinct mechanisms. Comparing the two screens, we found molecular and phenotypic evidence for similarities as well as differences between peripheral da and central RP2 neurons in the implementation of dendrite growth, branching and targeting.

The third aim of this study was to understand whether peripheral and central neurons are largely similar or different in the molecular mechanisms they employ for dendrite morphogenesis. Conservation of some mechanisms and factors required for dendrite morphogenesis have been shown to extend across neuronal cell types and phyla 7, including the cadherin Flamingo/Celsr 718384 and the guidance cue Slit and its receptor Robo 33348285. To ascertain the degree of similarity between peripheral da and central RP2 dendrite development, we determined the overlaps in the phenotypes recovered from the screens and the genes identified. Based on phenotypes, underlying cellular strategies clearly exist that are common (for example, growth and branching) as well as distinct (for example, targeting and ecdysone response). Furthermore, there was also a sizable overlap of 39% among the 62 candidate genes identified in both the da and RP2 screens, suggesting partially overlapping cytoplasmic and nuclear regulatory programs (Figure 9). However, at a more stringent level of comparison we found that particular GS lines do not necessarily induce comparable phenotypes in both cell types. In fact, candidate genes that were common to both screens were as likely to generate different phenotypes as similar ones in peripheral and central neurons. Only four genes common to both screens gave similar phenotypes in both screens (CG7518 (GSd211), CG2617 (GSd450), CG33298 (GSd454), and shn (GSd472). This observation suggests a more differentiated scenario, namely that the molecular implementation and regulation of dendrite morphogenesis in peripheral da and central RP2 neurons is largely cell type-specific, perhaps because they have distinct cellular requirements to accommodate differences in environmental complexity, the specificity of intercellular connections, or the integration of function and morphology. Nonetheless, the RP2 screen identified three genes not expressed in central neurons, but required for dendrite morphogenesis in peripheral (da) neurons (for example, ab, nos,ttk). We interpret this cross-fertilisation between the screens as an indication that certain cellular tasks required for the extension, branching and stabilisation of dendrites are supported by similar, though perhaps not identical, regulatory mechanisms in both cell types.

GAL4 driver lines were: en-GAL4 21; Tub84B-FRT-CD2-FRT-Gal4 86; GAL4^109(2)80 15; GAL4²²¹ 28; C161-GAL4 87; and ppk1.9-GAL4 17. UAS-lines were: UAS-mCD8::GFP 63; UAS-EcR-DN (UAS-EcR-B1^W650A) 65; UAS-IR-EcR 64; and UAS-Dcr2 88.

GS is a bidirectional UAS-based P-element that can activate genes on either side of the insertion site 2489. As described previously 23, the 141 independent GS lines screened here were pre-selected from a larger collection (1,127) for having lethal effects when misexpressed in the entire embryonic nervous system (using scrt^11–6-GAL4), so as to enrich (eight-fold) for genes likely to disrupt neuronal morphology or function.

Forty-four PNS neurons form per abdominal hemi-segment in three clusters 9091. We focused on the dorsal-most cluster in which there are eight neurons with multiple dendrites (md), including one tracheal dendrite (td) neuron, one neuron with bipolar dendrites (bd), and six dendritic arborisation (da) neurons. The pattern of dendrite outgrowth from dorsal cluster da neurons has been shown to be consistent from embryo to embryo 1578.

For screening in embryos, each of the 141 GS lines was crossed to the recombinant fly line GAL4^109(2)80, UAS-mCD8::GFP. Eggs were collected for 2 hours at 25°C, incubated at 25°C for another 17 hours, then shifted to 4°C for 24 hours to allow for improved visualization of the GFP signal. The resulting late stage 17 embryos were manually de-chorionated, fixed 5 minutes in 4% paraformaldehyde, immersed in halocarbon 200 oil (Halocarbon Products, River Edge, NJ, USA), coverslipped, and examined with confocal microscopy using a Yokogawa spinning disk confocal system (Perkin-Elmer, Waltham, MA, USA) on an Eclipse TE2000-U microscope (Nikon). Z-series images (60× objective) were collected using Metamorph software (MDS Inc., Mississauga, ON, Canada). Image stacks (approximately 15 optical sections, step size 0.2 μm) were exported to Photoshop software (Adobe) and prepared for publication by converting images to greyscale and adjusting brightness and contrast. Neurons from at least 20 individuals per genotype were examined.

For screening in larvae, wandering third instar larvae from crosses described above were collected just prior to pupation from uncrowded vials and dissected in 80% glycerol and 20% phosphate-buffered saline (PBS). A scalpel was used to excise the anterior end of each larva, allowing gut, fat body, and tracheal tubes to be removed. Dissected larvae were gently stretched, then squashed under a coverslip and imaged as for embryos, using a 25× objective to capture the entire dorsal cluster field or a 60× objective for detailed features of individual da neurons.

Maximum projections of captured confocal Z-series image stacks of ddaC neurons were compiled with Photoshop (Adobe), then imported into Reconstruct software 92 to count the number of branch ends per neuron, the length of primary branches (class III) as well as the area of each dendritic field (class IV) using the polygon method 25. Data were exported from Reconstruct and statistical analysis was performed using Analyse-It software for Microsoft Excel. The data were tested for normal distribution using the Shapiro-Wilk test, and the probability of unequal variance was calculated with the F-test. When only one experimental condition was compared with a control, a two-tailed t-test was performed. For multiple experimental groups, comparisons among all pairwise combinations were made using one-way ANOVA (Tukey).

The RN2-FLP transgene was generated by modification of the e5z3 plasmid generously provided by Miki Fujioka. The lacZ coding region was replaced with a 5' linker and the Flippase coding region so as to join it to a tandem repeat of the even-skipped (eve) +7.9–8.6 kb regulatory element, which confers expression to the aCC and RP2 motorneurons and, weakly, the pCC interneuron 93. Transgenic flies were generated by standard procedures. A recombinant third chromosome was generated in order to obtain a stock of the following genotype w^-;; RN2-FLP, Tub84B-FRT-CD2-FRT-GAL4, UAS-mCD8::GFP. Thus, expression of yeast Flippase in the RP2 precursor induces recombination between the FRT sites in a stochastic fashion, inducing expression of the UAS-mCD8::GFP reporter under Tub84B-GAL4 control in isolated RP2 (and occasionally aCC and pCC) neurons.

Each of the 141 GS lines was crossed to the FLPout recombinant stock. Following 6 hour egg collections at 25°C, agar collection plates were transferred to 29°C for a further 20 hours. Hatched larvae were put into a drop of PBS on a microscope slide and squashed under a coverslip until flat but still intact. A minimum of five animals with several labelled RP2 neurons each were examined for each GS line under widefield fluorescence using a Zeiss Axiophot microscope, a 60×/0.9 water dipping objective (Olympus) and Zeiss AxioCam MR controlled by AxioVision 4.1 software. Reproducibility of dendritic phenotypes was confirmed for most lines by repeat crosses and confocal analysis of dissected nerve cords.

For quantification ratios of anterior/posterior or medial/lateral dendrites, dendritic volumes were calculated using ImageJ 94: first, image stacks were split into either medial and lateral dendrites (defined by the central intermediate Fasciclin II fascicle) or anterior and posterior dendrites (the midpoint being the section of axon from which the primary dendrites emerge); second, pixels showing axon and cell body were removed manually from each focal plane; third, each focal plane was converted to a binary image after removal of background by applying the same threshold to all focal planes and all image sub-stacks (for example, of anterior and posterior dendrites); fourth, the total number of pixels in the stack was then calculated as a measure of dendritic volume. These values do not represent actual but relative dendritic volumes, which allow comparisons to be carried out between parts of the same dendritic tree. Numbers of branch points, lengths of dendritic segments and total dendritic tree lengths were calculated from dendritic trees that had been reconstructed using AMIRA software (Visage Imaging, Fuerth, Germany) and a customised reconstruction module by JF Evers 4243.

Nerve cords of first instar larvae were dissected in Soerensen phosphate buffer (pH 7.4), attached dorsal side up to Poly-L-Lysine (Sigma-Aldrich, St. Louis, MO, USA) coated coverslips and fixed in 2% paraformaldehyde, 3% sucrose in PBS for 60 minutes at room temperature. Washes were done in PBS with 0.3% Triton-X-100 (Sigma), 5 changes over the course of 60 minutes. Embryos and third instar larvae were dissected in PBS, fixed for 10 minutes in 4% paraformaldehyde, then washed for 2 hours at room temperature. Specimens were incubated in primary antibodies overnight at 4°C, washed, incubated in secondary antibodies either for 2 hours at room temperature or overnight at 4°C. In some cases, the muscles overlying the dorsal cluster PNS neurons were removed for better visualization of da neurons. Specimens were mounted in Vectashield (Vector Laboratories, Burlingame, CA, USA) and imaged with a SP1 point scanning confocal system (Leica) or a CSU-22 spinning disk confocal system (Yokagawa), operated with MetaMorph (Molecular Devices).

Primary antibodies were: anti-Fasciclin II (MAb 1D4) supplied by the Developmental Studies Hybridoma Bank (DSHB; 1:20 dilution), goat anti-GFP (1:1,500; Abcam, Cambridge, MA, USA) and mouse anti-EcR antibodies (1:10 dilution; DSHB). The mouse anti-EcR antibodies included: all EcR isoforms (mAb Ag10.2); EcR-A (mAb 15G1a); or EcR-B1 (mAb AD4.4). Primary antibodies were diluted in PBS, 0.3% Triton-X-100. Secondary antibodies were Alexa fluorophore (Alexa488, Alexa561 and Alexa633; Invitrogen, Carlsbad, CA, USA) and Rhodamine Red-X conjugated secondaries (Jackson ImmunoResearch, West Grove, PA, USA), highly cross-absorbed applied at a 1:800 dilution in PBS, 0.3% Triton-X-100.

In situ hybridization was performed as described in 95 with 0.3% SDS present in the hybridisation solution. Probes were generated from cDNA clones (Drosophila Genomics Resource Center) using the Megascript kit (Ambion/Applied Biosystems, Austin, TX, USA) and DIG-UTP (Roche, Basel, Switzerland) with the following primers. pOT2: 5', ATTAACCCTCACTAAAGGGAGCAGATCTGATATCATCGCCACT; 3', TAATACGACTCACTATAGGGAGAACGCGGCTACAATTAATACATAACC. pBSK^-: 5', GCGCGCAATTAACCCTCACTAAAGGG; 3', GCGCGCGTAATACGACTCACTATAGGG. pOTB7: 5', ATTAACCCTCACTAAAGGGACTAAGGTAGCGAGGCCTGGGTGG; 3', GCGCGCAATTAACCCTCACTAAAGGG.

To characterize the mutant phenotype of usp in da neurons, flies of the stock FRT19A, tub-GAL80, hs-FLP; GAL4^109(2)80, UAS-mCD8::GFP were crossed to flies from the following stocks: 1. FRT19A, w⁺ 2. FRT19A, UAS-mCD8::GFP, usp²; λ10,Tb/TM3 3. FRT19A, usp³/FM7c and 4. FRT19A, usp⁵/FM7c. Embryos were collected for 2 hours, incubated at 25°C for 2–3 hours, then heat-shocked at 38°C for 1 hour and incubated at 25°C until they were analyzed as wandering third instar larvae, just prior to pupation.