To examine ewg mutants for synaptic growth defects at third instar NMJs, we used a clonal analysis strategy in mosaic animals. For this analysis we made the following rescue construct, termed eFeG. The ewg cDNA was flanked by FRT sites and fused to an elav promoter. To visualize recombination, the sequence of yeast GAL4 was inserted downstream of the FRT-ewg-FRT cassette. FLP/FRT mediated recombination will result in loss of the ewg cDNA and lead to expression of GAL4 in neurons that can be visualized in the presence of a UAS-CD8::GFP transgene (Figure 1a,c,e-i) 172425. The functionality of eFeG is shown in a clone induced in photoreceptor neurons in ewg^l1, a null allele 24. Upon loss of the ewg cDNA in the clone, CD8::GFP is expressed and EWG expression is lost (Figure 1b,c). For the analysis of third instar NMJs, recombination was induced in late embryogensis (14-16 h after egg laying). Larvae bearing EWG null clones move indistinguishably from their balancer carrying siblings, pupate normally and many adults hatch. These adults are, however, impaired in walking.

At third instar NMJs, EWG deficient motorneurons have an increased number of synaptic boutons that look morphologically normal as visualized with an antibody against synaptotagmin, a marker for synaptic vesicles 26 (Figure 1d-i). Quantification of type 1b boutons at muscle 13 revealed about an 85% increase in bouton number (p < 0.0001; Figure 1j) compared to wild-type or balancer carrying siblings containing one wild-type copy of ewg (rec. control in Figure 1j). Since ewg is located on the X chromosome and we used males, we also quantified type 1b bouton numbers at muscle 13 in ewg^l1females transheterozygous for a hypomorphic P element insertion, P{lacW}ewg^G1518 (18.8 ± 0.45, n = 18, p < 0.0001 for comparison with wild type). Significant effects were also observed at muscles 4, 6/7 and 12 (29%, 30% and 45% increase compared to wild type, p ≤ 0.001). Besides overgrowth, NMJs of EWG deficient clones appeared normal and staining with markers for active (Nc82) and periactive (Highwire) zones, for microtubules (Mab 22C10) or post-synaptic specializations (anti-DLG) did not reveal obvious differences (Figure 1k-n and data not shown). The role of EWG in synaptic growth regulation is cell-autonomous, since bouton numbers of non-recombined neurons in mosaic animals were indistinguishable from wild type (data not shown). Inhibition of synaptic transmission by expression of tetanus toxin in EWG deficient neurons did not affect overgrowth, demonstrating that overgrowth is not a result of compensatory signals from the muscle due to changes in neuronal activity (Figure 1j). Similarly, expression of tetanus toxin also did not affect synaptic growth in wild-type clones (Figure 1j).

To demonstrate that synaptic overgrowth of EWG deficient neurons is a result of ewg loss of function (LOF), we added a rescue construct where an elav promoter drives the ewg cDNA (elav-EWG) 17. Presence of this construct significantly reduced the number of type 1b boutons at muscle 13 compared to EWG deficient neurons (p < 0.0001; Figure 1f), but rescue is not complete compared to wild-type neurons (p < 0.001; Figure 1f). Since ewg is strongly homologous to human NRF-1 (80% in the DNA binding domain 18), we tested if expression of NRF-1 under an elav promoter can rescue the synaptic growth defect of ewg. elav-NRF-1 transgenes fully rescued synaptic growth defects of ewg^l1mutants (p < 0.0001), resulting in bouton numbers that were not significantly different from wild type (Figure 1j). elav-NRF-1 transgenes also fully rescued embryonic lethality of ewg^l1mutant embryos (89 ± 8%, n = 5 independent transgene insertions). ewg^l1elav-NRF-1 animals develop to pharate adults, but fail to hatch.

Since ewg LOF results in synaptic overgrowth, we wanted to know if overexpression of ewg results in reduced synaptic growth. To separate ewg function in early neuronal differentiation from its role in synaptic growth regulation and to apply physiological concentrations of EWG, we used an inducible neuron-specific GAL4 driver (elav-GeneSwitch-GAL4, elav-GS-GAL4 27) to express EWG from a UAS-ewg transgene. Concentration dependent inducibility of the elav-GS-GAL4 was verified by western blots (Figure 2a). At NMJs, numbers of type 1b boutons at muscle 13 were about 30% reduced compared to wild type, and at muscle 12 about 45% reduced (p < 0.0001; Figure 2b-d) when EWG expression was about two-fold increased compared to wild type (Figure 2a, lane 3 versus lane 2). Higher expression levels of EWG did not further reduce synaptic growth, but resulted in pupal lethality. Controls with induced elav-GS-GAL4 (but no UAS-ewg), or UAS-ewg alone had wild-type bouton numbers (data not shown).

In the presence of the RNA binding protein ELAV, the last ewg intron is spliced, resulting in expression of EWG protein 1921. Since the ewg gene encodes a transcription factor, EWG could potentially regulate a large portion of genes that are also regulated by ELAV, and therefore rescue elav mutants. To test if elav mutants are rescued by EWG, we used animals transheterozygous for the elav^ts1temperature sensitive allele and for the null allele elav^e528. When early functions of ELAV in neuronal differentiation were allowed by rearing embryos at the permissive temperature, elav-EWG fully rescued the lethality of elav^ts1/elav^e5flies (Figure 3a). The rescued elav^ts1/elav^e5; elav-EWG animals, however, showed motor defects and were flightless, suggesting that the RNA binding protein ELAV regulates additional genes. Given the prominent NMJ phenotype of ewg mutants, we next tested if ELAV is also involved in regulating synaptic growth. At NMJs, elav^ts1/elav^e5animals showed a significant reduction of bouton numbers compared to wild type (p < 0.0001; Figure 3c,f), a phenotype opposite to the ewg mutant phenotype. Synaptic growth defects of elav^ts1/elav^e5animals, however, were fully rescued to wild-type levels by an elav-EWG transgene (p < 0.0001 for elav^ts1/elav^e5; elav-EWG compared with elav^ts1/elav^e5and not significantly different from wild type; Figure 3d,f). As expected, an elav-ELAV transgene also fully rescued synaptic growth defects of elav^ts1/elav^e5animals (p < 0.0001 for elav^ts1/elav^e5; elav-ELAV compared with elav^ts1/elav^e5and not significantly different from wild type; Figure 3d,f). These results indicate overlapping stimulatory and restrictive pathways in regulating synaptic growth that are integrated through the transcriptional regulator EWG (see below). Boutons in elav^ts1/elav^e5animals appeared normal and staining with markers for active (Nc82) and periactive (Highwire) zones, for microtubules (Mab 22C10) or post-synaptic specializations (anti-DLG) did not reveal obvious differences (data not shown).

To assess how the transcription factor EWG regulates synaptic growth, we identified genes differentially regulated in ewg^l1mutants using cDNA microarrays. We hand-selected ewg^l1late stage embryos that differ from their siblings by the lack of green fluorescent protein (GFP), extracted and amplified polyA RNA, and hybridized cDNA microarrays. To exclude genes differentially regulated due to genetic background and to validate genes differentially regulated in ewg^l1mutants at a genomic scale, we included ewg^l1animals rescued by elav-EWG. From these experiments (n = 4, except for ewg^l1n = 5), mRNA expression levels of 107 genes were significantly down-regulated (Figure 4c) and of 107 genes were significantly up-regulated (Figure 4d) in ewg^l1embryos (p ≤ 0.01; changes in expression levels are shown in Table S3 of Additional data file 1) compared to both wild-type and ewg^l1elav-EWG embryos. Quantitative RT-PCR analysis from selected genes that were down-regulated (Ac3, CG7646, CG1909, Srca) or up-regulated (Pepck, osi14, CG5171, CG10585) in ewg^l1mutant embryos further confirmed the findings from the microarray expression analysis (Figure 4e), for example, expression levels were reduced or increased in ewg^l1mutant embryos compared to the control gene elav, but restored to wild-type levels in the presence of the elav-EWG transgene.

Clustering differentially regulated genes revealed that down-regulated genes fall into several functional classes with a preference for genes involved in the regulation of gene expression (40%, 43 of 107; Figure 4a,c), while up-regulated genes are enriched for genes involved in basic cellular metabolism (47%, 50 of 107; Figure 4b,d). Surprisingly, only a minor portion of genes (4%, 8 of 214; Figure 4c,d) could be loosely defined as 'neuronal effector genes' (genes at the end of the regulatory expression hierarchy in neurons and associated with neuron specific functions) with known or potential functions in synaptic growth regulation, for example, Ac3, encoding an adenylyl cyclase, CG1909, encoding a Drosophila Rapsyn homologue, CG7646, encoding a Ca²⁺ sensor, and the gene encoding Henna, a phenylalanine hydroxylase involved in dopamine and serotonin synthesis. Candidates for mediating the synaptic growth phenotype from the remaining functional classes include groucho (gro), encoding a transcriptional repressor of Wg and N signaling, and genes encoding a number of cell adhesion molecules (CG7227, 18w, CG8434, CG7896, Gli, CG4115).

Synaptic growth defects in ewg mutants are cell-autonomous, suggesting that differentially expressed genes involved in regulating synaptic growth are also expressed in the nervous system. We therefore analyzed RNA expression patterns in late stage embryos from a representative number of genes differentially regulated in ewg^l1mutants (66%, 74 down- and 68 up-regulated genes; RNA in situ expression patterns are mostly available from the Berkeley Drosophila Genome Project (BDGP)). RNA in situ hybridization experiments revealed that 86% (42 of 52) of down-regulated genes in ewg^l1mutants, but only 17% (11 of 63) of up-regulated genes in ewg^l1mutants, are expressed in the nervous system of late stage embryos (Figure 4c,d; Figures S2 and S3 in Additional data file 1). Interestingly, these RNA in situ hybridization experiments further revealed that the vast majority of genes down-regulated in ewg^l1mutants are broadly expressed in the entire nervous system, suggesting that they might be direct targets of EWG. From those genes where the RNA expression pattern was analyzed, 30% (22 of 74) of down-regulated genes in ewg^l1mutants and 7% (5 of 68) of up-regulated genes in ewg^l1mutants did not yield a signal in RNA in situ experiments, although they were detected on microarrays at comparable levels. These transcripts might reside, therefore, in tightly packed mRNP particles and not be accessible to in situ hybridization after fixation, which has been shown, for example, for Fragile X Mental Retardation Protein (FMRP) mRNA in dendrites 29.

Recent efforts in Drosophila genome projects have increased the number of mutants to cover 50-60% of all protein coding genes 3031323334. This allows for a functional genomics approach to identify a representative fraction of genes differentially regulated in ewg^l1mutants and involved in synaptic growth regulation, and to elucidate general principles that operate in this biological process. We obtained viable mutants for 42% (40 of 95) of genes down-regulated in ewg^l1mutants and for 37% (23 of 62) of genes up-regulated in ewg^l1mutants (Table S2 in Additional data file 1). Focusing on synaptic growth regulation, we restricted our analysis to genes not involved in basic cellular metabolism unless their expression was enriched in the nervous system. Most of the mutants were P element or piggyBAC insertions in promotor regions and represent hypomorphic alleles indicated by 38% (21 from 57) showing adult phenotypes (pupal lethality, morphological aberrations, sterility or flightlessness; Table S2 in Additional data file 1) as transheterozygotes over a deficiency or a second allele. A minor portion of mutants was embryonic or early larval lethal (10%, 7 of 70) and could not be analyzed for synaptic growth defects. Viable mutants tend to accumulate genetic modifiers 35. Therefore, we normalized the genetic background by analyzing transheterozygotes for a chromosomal deficiency, or in some cases transheterozygotes for a second allele. Since the most pronounced effect in ewg^l1mutants was at muscle 13, we quantified type 1b boutons at muscle 13.

Analysis of synaptic growth in mutants of genes down- and up-regulated in ewg^l1mutants revealed that 85% (34 of 40) and 70% (16 of 23), respectively, were associated with statistically significant differences in synaptic growth compared to controls (Figure 5, y w line and y w transheterozygous for appropriate deficiencies, p ≤ 0.0001 for the vast majority; for details, see Table S2 in Additional data file 1). Ninety percent (18 of 20) of the genes down-regulated in ewg^l1mutants are also expressed in the nervous system (Figure 5a,c). Clustering these genes into functional classes revealed a strong enrichment for genes involved in the regulation of gene expression (65%, 22 of 34; Figure 5a,c). In contrast, genes up-regulated in ewg^l1mutants with synaptic growth defects are mostly not expressed in the nervous system (64%, 7 of 11; Figure 5b,c) and are split into many different functional classes (Figure 5b,c).

For both down- and up-regulated genes in ewg^l1mutants with synaptic growth defects, phenotypes could be split into groups with either more or less boutons (Figure 5). Among those genes down-regulated in ewg^l1mutants with growth defects in mutants, 29% (10 of 34) had the same overgrowth phenotype as ewg^l1mutants (Figure 5a). These results show that EWG-regulated gene expression combines both stimulatory and restrictive functions in synaptic growth regulation, and that restrictive functions dominate, indicated by the overgrowth phenotype of ewg^l1mutants.

Based on synaptic growth phenotypes associated with mutants of genes differentially regulated in ewg^l1mutants, the following model assigns more distinct roles to these genes in the context of EWG regulation (Figure 5c). Genes down-regulated in ewg^l1mutants with an increased number of boutons exert restriction on synaptic growth in the presence of EWG, while those genes down-regulated in ewg^l1mutants with a reduced number of boutons in mutants might provide a trophic supply for synaptic growth. For genes up-regulated in ewg^l1mutants, overgrowth associated with mutations in these genes is indicative of a role in expansion of synaptic growth as they are repressed in the presence of EWG. This class includes primarily cell adhesion molecules for which a restrictive role has been demonstrated when adhesion has been increased (for example, 36). Most genes up-regulated in ewg^l1mutants with fewer boutons are not expressed in the nervous system. These genes might, therefore, be regulated endocrinologically, indicated by the role of ewg in regulating metabolic aspects.

Pioneering work in yeast has shown that expression of functionally related genes is co-regulated (reviewed in 37). We therefore used genetic interaction experiments to test for functional connections among genes that are differentially expressed in ewg mutants and that are also involved in synaptic growth regulation. Since loss of ewg results in more boutons and synaptic overgrowth can be further induced (for example, by overexpression of the fos and jun heterodimer AP-1, see below), we analyzed double mutants in a representative number of those genes down-regulated in ewg^l1mutants with more boutons. For all different combinations tested, none of the double mutant animals had more boutons than the single mutants, indicating that these genes do not act in parallel to regulate synaptic growth as independent effects would be additive (Table S1 in Additional data file 1). Some combinations, however, resulted in significant reduction of bouton numbers (for example,Ac3; Bcl7-like, Ac3; CG1943, and Bcl7-like; CG12299; Table S1 in Additional data file 1), suggesting that a combined loss of function in these genes might affect trophic supply to synaptic growth.

Next, we validated functional connections among genes down-regulated in ewg^l1mutants with a synaptic overgrowth phenotype in another assay. One of the genes in this class is gro, for which hypomorphic alleles are known that are associated with an overproliferation of frontal bristles on the head (Figure 6b). gro has been described as a transcriptional co-repressor that interacts with a subset of negative transcriptional regulators 38. We therefore tested a representative number of mutants in genes downregulated in ewg^l1mutants in combination with gro for a change of the gro bristle phenotype. All mutants tested genetically interacted with gro, resulting in either an enhancement or suppression of the gro bristle phenotype (Figure 6e). None of the single mutants had more bristles, but some had less (Figure 6d), suggesting that the gro phenotype can be suppressed (for example, CG8924 and Bcl7-like), indicated by several roles of gro in peripheral nervous system specification 39. With quantitative RT-PCR, we verified differential expression of these genes in ewg^l1mutants and rescue to wild-type expression levels by the presence of an elav-EWG transgene in ewg^l1mutants (Figure 6f). The genes are also expressed predominantly in the ventral nerve cord (Figure S3 in Additional data file 1; Figure 4c) 40. Taken together, these data strongly suggest that these genes down-regulated in ewg^l1mutants are functionally connected and operate in a common pathway.

The predominance of transcriptional and post-transcriptional regulators among genes differentially regulated in ewg^l1mutants, together with opposite phenotypes associated with mutants of these genes, suggests that overlapping stimulatory and restrictive pathways are integrated by an extensive regulatory gene network. This model for the regulation of synaptic growth implies that signaling pathways converge in the regulation of gene expression and predicts that ewg genetically interacts with several signaling pathways involved in regulating synaptic growth.

Groucho, the protein encoded by gro, is differentially regulated in ewg mutants and acts in both N and Wg signaling pathways 394142. Since Wg was previously shown to regulate synaptic growth 43, we first determined if N is also involved in regulating synaptic growth and second, if ewg genetically interacts with these two pathways. Therefore, we quantified type Ib bouton numbers at muscle 13 of third instar NMJs in the absence or presence of EWG using the eFeG transgene (Figure 1). In this genetic condition, half of the mosaic animals will have one copy of the wild-type ewg gene in clones while the other half contains the ewg^l1null allele and, therefore, has no EWG protein in clones. Changes in N signaling were achieved through the recombined eFeG transgene that expresses GAL4 in the clone and drives either UAS-N for N overexpression or UAS-N-RNAi for N down-regulation. Expression of UAS-N-RNAi in post-mitotic neurons has previously been shown to reduce N levels and to result in long-term memory deficits 111213. Changes in wg signaling in clones was achieved through overexpression from UAS transgenes of wg or pangolin (pan), the transcription factor intracellularly mediating the canonical wingless signal in embryos and wing discs 444546.

Similar to ewg, N restricts synaptic growth, resulting in increased bouton numbers indicated by down-regulation of N with RNA interference during larval life (from two independent UAS inserts, p = 0.005 for bar 1 compared with bar 4, Figure 7a) or in reduced bouton numbers when N was overexpressed (p < 0.0001 for bar 1 compared with bar 6, Figure 7a) compared to wild-type control.

In the absence of EWG, removing N (N LOF) resulted in intermediate numbers of boutons compared to wild type and ewg LOF (p < 0.0001 for bar 5 compared with bar 1 and 2, Figure 7a). The increase in bouton numbers in N LOF (bar 5 compared with bar 2) is not additive to the numbers observed in the absence of EWG (bar 2 compared to bar 1, as seen for AP-1 overexpression, see below), suggesting that regulation of synaptic growth by N is not independent of ewg and indicating that part of the N signal is required for the full effect seen in ewg LOF. Overexpression of N (N GOF) in the absence of EWG resulted in intermediate numbers of boutons compared to wild type and ewg LOF (p < 0.0001 for bar 7 compared with bars 1 and 2, Figure 7a).

Overexpression of wg increased bouton numbers (p < 0.0001 for bar 1 compared with bar 8, Figure 7b), as previously observed 43. Overexpression of pan also resulted in increased bouton numbers (p < 0.0001 for bar 1 compared with bar 10, Figure 7b), demonstrating that the canonical wg pathway operates presynaptically to stimulate synaptic growth.

In the absence of EWG, overexpression of both wg and pan were epistatic to ewg LOF and bouton numbers were significantly reduced compared to ewg LOF (p ≤ 0.0001 for bar 2 compared with bars 9 and 11, Figure 7b).

The fos and jun heterodimer AP-1, and TGF-β signaling comprise two other known pathways involved in regulating synaptic growth 4578. Both AP-1 and TGF-β stimulate synaptic growth compared to wild type as evident either by overexpression of fos and jun together (p < 0.0001 for bar 1 compared with bar 12, Figure 7c) and an activated BMP type I receptor (tkvA, from two copies of UAS-tkvA, p < 0.0001 for bar 1 compared with bar 16, Figure 7d), or by removal of AP-1 activity through overexpression of a dominant negative form of jun (junBZ, p < 0.0001 for bar 1 compared with bar 14, Figure 7c), or in the BMP type II receptor mutant wishful thinking (wit, p < 0.0001 for bar 1 compared with bar 18, Figure 7d), which is largely in agreement with previous observations 4578. Overexpression of either fos or jun alone has no effect on synaptic growth 4 (and data not shown).

Next, we wanted to define how these two pathways relate to ewg mediated regulation of synaptic growth. In the absence of EWG, overexpression of AP-1 is additive and further increases bouton numbers significantly (p < 0.0001 for bar 13 compared with bars 2 and 12, Figure 7c). Inhibiting AP-1 function by expressing the dominant negative junBZ completely removes the stimulatory effect of ewg LOF (p < 0.0001 for bar 15 compared with bars 2 and 13, and non-significant for bar 15 compared to bar 14, Figure 7c), suggesting that AP-1 functions downstream of ewg.

Overexpression of tkvA (from two copies of UAS-tkvA) in ewg LOF resulted in intermediate numbers of boutons compared to wild type and ewg LOF (p < 0.0001 for bar 17 compared with bars 1 and 2, Figure 7d), indicating that the TGF-β signaling pathway does not act independent of ewg in synaptic growth regulation. Removal of the BMP type II receptor wit significantly reduced bouton numbers in ewg LOF (p < 0.0001 for bar 19 compared with bars 2 and 17, Figure 7d). The strong reduction of bouton numbers in wit mutants in ewg LOF compared to wild type (p < 0.0001 for bar 19 compared with bar 1, Figure 7d) further suggests that TGF-β signaling also includes a component that acts downstream of ewg.

Fly breeding, genetics and P-element mediated transformation of Drosophila, and recombinant DNA technology were done according to standard procedures as detailed in Soller and White 21 and Soller et al. 65. In all experiments the ewg NS isoform was used 24. The eFeG construct P{w+ = elav-FRT-EWG-FRT-GAL4} is basically the same as the EWG^NS construct 25 except that the ewg cDNA is flanked by FRT sites followed by the coding sequence of the yeast GAL4 transcription factor.

For cloning of the eFeG construct, the 3.5 kb elav promoter 25 was cloned into a linker modified CaSpeR 4 (Drosophila Genomics Resource Center, Bloomington, IN, USA) with a blunt EcoRI site and NotI site to generate the construct C4MMelav2. The 5' FRT site was added to the ewg gene by nested PCR with primers FRTewgF1 (AGAAAGTATAGGAACTTCAGAGCGCTTTTGAA

GCTAgc

AATTC

For the analysis of synaptic growth in EWG deficient neurons, two independent inserts of the eFeG construct on the X and third chromosomes that fully rescue ewg^l1mutants were used. To induce clones with eFeG transgenes, 14-18 h old embryos from ewg^l1eFeG females crossed with males carrying hs-flp UAS-CD8::GFP (P{hsFLP}38, P{UAS-mCD8::GFP.L}LL5 6667) on the second chromosome were heat shocked for 45 minutes at 35°C, rested for 45 minutes at room temperature and incubated for another 45 minutes at 37°C. For genetic interaction experiments, the ewg^l1eFeG and hs-flp UAS-CD8::GFP chromosomes were combined with the following UAS transgenes or mutants (from Bloomington stock center, Bloomington, IN, USA) or as noted: UAS-wg (second or third, P{w [+mC] = UAS-wg.H.T:HA1}6C and P{w [+mC] = UAS-wg.H.T:HA1}3C 45), UAS-pan (second or third, P{w [+mC] = UAS-pan.dTCF}24 and P{w [+mC] = UAS-pan.dTCF}4 45), UAS-N-RNAi_i (second and/or third, P{w [+mC] = UAS-N.dsRNA.P}14A and P{w [+mC] = UAS-N.dsRNA.P}9G 13), UAS-N (second and/or third, gift form S Artavanis-Tsakonas 68), UAS-tkvA (second and/or third, gift from M O'Connor 7), UAS-junBZ (second or third, P{Jbz}1 and P{Jbz}10 69), UAS-fos (P{UAS-Fra}, second 69), UAS-jun (P{UAS-Jra}2, third 69) and wit^A12 with wit^B11 8, UAS-Tetanustoxin (gift from S Sweeney, P{UAS-TeTxLC.tnt}, CNT-E, second 70). Overexpression of UAS-wg, UAS-pan, UAS-N-RNAi and UAS-N with elav-GS-GAL4 (third, P{elav-Switch.O} 27) resulted in significant changes in bouton numbers comparable with overexpression with recombined eFeG compared to wild type.

The elav-NRF-1 construct was made by exchanging the ewg open reading frame (ORF) in elav-EWG with the ORF of NRF-1 as follows. The NRF-1 ORF was amplified by PCR with primers NRF-F1 (TAGA

GCGGCCGCTCGAGAATTC

Embryos for microarray analysis were produced by crossing ewg^l1/C155-GAL4; UAS-GFP/+ females to C155-GAL4; UAS-GFP males (P{GawB}elav^C155; P{UAS-GFP.S65T}T2 6573). The X-chromosomal C155-GAL4 enhancer trap was inserted in the neighboring elav gene and drives GFP expression in neurons and neuroblast 65. From the progeny of this cross, non-GFP expressing embryos, which are male, were then hand-picked 16-18 h after egg laying and RNA was extracted from pools of 50 embryos with Trizol (Invitrogen, Carlsbad, CA, USA). Pooled RNA from a total of 300 embryos was then reverse-transcribed with an oligo-dT primer containing a T7 promoter and, after the synthesis of the second strand, linearly amplified by in vitro transcription with T7 RNA polymerase and labeled with biotinylated NTPs according to the manufacturers instructions (Genechip Protocol for Eukaryotic Target preparation, Affymetrix, Santa Clara, CA, USA) yielding between 20 and 50 μg of RNA. Amplified RNA (15 μg) including spike controls was hybridized to Affymetrix Drosophila genome arrays Version X using the Affymetrix Gene Chip Instrument System overnight at 42°C. Arrays were washed and stained with streptavidin-phycoerythrin before scanning on an Affymetrix Gene Chip scanner.

Quantitative RT-PCR was done with 5' ³²P labeled forward primers and PCR products from unsaturated cycles (24, 26, 28 or 30 depending on expression levels of a particular gene) were analyzed on 6% polyacrylamide gels essentially as previously described 19. Primer sequences are available upon request.

Gene chip data were analyzed using the software packages MAS version 5 (Affymetrix) and dChip. Scanned data were first processed with MAS to convert raw image files (.DAT) to probe signal values files (.Cel). Probe signal files were normalized across samples using dChip invariant set method. Summary values for each probe set were calculated with PM-only model in dChip. Microarray data have been deposited in MIAMExpress 74 under the accession number E-MEXP-1312.

Western blotting was done as previously described 24. Quantification of western blots was done with Quantity ONE 4.2.3 (Bio-Rad, Hercules, CA, USA). RNA in situ hybridizations were done according to the BDGP protocol, or obtained from the BDGP web page 75. For the analysis of NMJs, wandering third instar larvae were dissected in phosphate-buffered saline and fixed with 4% formaldehyde for 30 minutes. The following antibodies were used: anti-DSYT (1:200; gift from T Littelton 26), anti-DLG (1:100; gift from V Budnik 76), anti-CSP (1:100; gift from K Zinsmaier 77), anti-Nc82 (1:6; Developmental Studies Hybridoma Bank), anti-Highwire (1:6; Developmental Studies Hybridoma Bank), Mab22C10 (1:200; Developmental Studies Hybridoma Bank), FITC conjugated anti-HRP (1:200; Cappel, Cochranville, PA, USA), FITC conjugated anti-CD8 (1:200; Cappel). FITC, TRITC (Jackson Labs, Bar Harbor, MA, USA) and Cascade Blue (Molecular Probes, Eugene, OR, USA) conjugated secondary antibodies were used 1:200. Confocal images were acquired at 18°C using a Leica TCS SP2 confocal scanning microscope with a Plan APO HC 10 × 0.4 numerical aperture objective for eye discs, or under oil with a Plan APO HCX 63 × 1.4 numerical aperture (NMJs) or a Plan APO HCX 100 × 1.4 numerical aperture (single boutons) objective and Leica software. Image files were converted to TIFF and merged using Photoshop CS2 (Adobe). Levels of individual channels were adjusted where applicable to maximize pixel range. Flies were photographed with a digital camera (Nikon) on a Leica Stereoscope at 3× magnification. Statistical analysis of bouton numbers was done with ANOVA followed by post hoc analysis with Fisher's PLSD (protected least significant difference) for multiple comparisons, or with t-test for two samples, using StatView.