Transcription factor codes play an essential role in neuronal specification and axonal guidance in both vertebrate and invertebrate organisms. However, how transcription codes regulate axon pathfinding remains poorly understood. One such code defined by the homeodomain transcription factor Even-skipped (Eve) and by the GATA 2/3 homologue Grain (Grn) is specifically required for motor axon projection towards dorsal muscles in Drosophila. Using different mutant combinations, we present genetic evidence that both Grn and Eve are in the same pathway as Unc-5 in dorsal motoneurons (dMNs). In grn mutants, in which dMNs fail to reach their muscle targets, dMNs show significantly reduced levels of unc-5 mRNA expression and this phenotype can be partially rescued by the reintroduction of unc-5. We also show that both eve and grn are required independently to induce expression of unc-5 in dMNs. Reconstitution of the eve-grn transcriptional code of a dMN in dMP2 neurons, which do not project to lateral muscles in Drosophila, is able to reprogramme those cells accordingly; they robustly express unc-5 and project towards the muscle field as dMNs. Each transcription factor can independently induce unc-5 expression but unc-5 expression is more robust when both factors are expressed together. Furthermore, dMP2 exit is dependent on the level of unc-5 induced by eve and grn. Taken together, our data strongly suggests that the eve-grn transcriptional code controls axon guidance, in part, by regulating the level of unc-5 expression.
Transcriptional codes within a neuron direct the expression of an array of receptors and adhesion molecules that will eventually determine its guidance and connectivity properties. A considerable number of guidance systems that control axon pathfinding and target selection have been identified in the last two decades (Kolodkin and Tessier-Lavigne, 2010). Nevertheless, how different transcriptional codes regulate the expression of those guidance systems is only starting to emerge.
Some reports have already established some links between transcription factor codes regulating axon guidance and the effector molecules they regulate in either motoneurons (Kania and Jessell, 2003; Broihier et al., 2004; Labrador et al., 2005; Shirasaki et al., 2006; Luria et al., 2008) or commissural neurons (Crowner et al., 2002; Pak et al., 2004; García-Frigola et al., 2008; Lee et al., 2008; Wilson et al., 2008). However, how this regulation ultimately leads to specific guidance decisions is still an open question.
The Drosophila neuromuscular system is composed of 36 motoneurons innervating a stereotyped array of 30 somatic muscles. In this system, two main nerve branches can be identified, the intersegmental nerve (ISN) and the segmental nerve (SN). Two motoneurons, aCC and RP2, within the ISN specifically project to the most dorsal muscles and a transcriptional code determined by the homeodomain transcription factor Even-skipped (Eve) (Doe et al., 1988; Landgraf et al., 1999; Fujioka et al., 2003) and the GATA2/3 homologue Grain (Grn) determines the specific guidance characteristics of these Drosophila dorsal motoneurons (dMNs) (Garces and Thor, 2006). The eve-grn code is the only one identified that has been dissected to single motoneuron resolution and Garces and Thor have shown that both transcription factors can work in parallel or sequentially in a cell-specific manner (Garces and Thor, 2006) making dMNs an ideal system to study transcriptional code regulation of axon guidance. Nevertheless, the target guidance systems they control in each cell or whether they regulate independent or common guidance systems is unknown.
To gain a better understanding of how the expression of receptors on the cell surface is regulated through transcriptional codes, we have studied Unc-5 receptor regulation in dorsal motoneurons. We show that grn regulates Unc-5 receptor expression in dMNs. Through genetic analysis we show that both grn and unc-5 function in the same pathway with grn working upstream of unc-5 to promote guidance of dMNs. We also show that both eve and grn are required independently to induce expression of unc-5 in dMNs and identify an enhancer element in the unc-5 locus responsible for expression in dMNs that responds to eve. Furthermore, grn and eve can promote unc-5 expression when misexpressed in dMP2 neurons. However, only the combined misexpression of the eve-grn dMN transcriptional code in dMP2 neurons can reprogramme their axonal pathfinding and promote axonal exit towards the muscle field (Garces and Thor, 2006). We now demonstrate that axonal exit of these ‘reprogrammed’ dMP2 neurons is directly dependent on the levels of unc-5 that both factors are able to induce together.
MATERIALS AND METHODS
The following stocks were used in misexpression experiments: UAS-grn#2;UAS-mEGFPF, UAS-grn#2,UAS-eve;UAS-mEGFPF, UAS-eve/TM3 (Brown and Castelli-Gair Hombría, 2000; Garces and Thor, 2006), UAS-HA-unc5 (Keleman and Dickson, 2001) and dMP2Gal4/CyO (Miguel-Aliaga and Thor, 2004). RN2Gal4; grnSPJ9/TM3 and UAS-HA-unc5;grn7L12/TM3 stocks were used in rescue experiments. The following stocks were also used: unc-58/CyO,XTE-18/CyO (unc-5 deficiency) (Labrador et al., 2005), Df(2R)eve, RP2A/CyO; RN2-Gal4,UAS-tau_lacZ (Fujioka et al., 2003), RN2-Gal4, UAS-tau_lacZ; grn7L12/TM3, dMP2Gal4,XTE18/CyO;UAS-tau-myc/TM2, Unc58/CyO;grn7L12/TM3, dMP2Gal4/CyO;UAS-tau-myc/TM2, Unc58,UAS-grn#2,UAS-eve, unc-5-Gal4 (Bloomington 47230), Df(2R)eve, RP2A/CyO; UAS-NLS-Venus and Df(2R)eve, RP2A/CyO; unc-5-Gal4.
Embryos were collected for 24 hours and processed as previously described (Labrador et al., 2005). ISN motor axonal projections at embryonic stage 16/17 in A2-A6 abdominal hemisegments were quantified in different genetic backgrounds. The following antibodies were used: anti-c-Myc 9E10 (1:50), anti-Fas2 1D4 (1:50) (Developmental Studies Hybridoma Bank), anti-HA (Covance; 1:500) and rabbit anti-β-gal (Cappel; 1:5000). Secondary antibodies were purchased conjugated to Alexafluor488 (Jackson ImmunoResearch Laboratories) and Cy3 (Molecular Probes) and used at 1:1000 and 1:500, respectively. Double-labelled images were false coloured for the benefit of colour-blind readers. Stacks of images were obtained with Zeiss Confocal LSM700 Microscope and a 40× oil immersion objective was used.
mRNA in situ hybridization
unc-5 in situ hybridization in aCC and RP2 dorsal motor neurons as well as dMP2 interneurons was performed with digoxigenin-labelled probes as previously described (Labrador et al., 2005). Hybridized probes were detected with anti-digoxigenin HRP, and Cy3-labelled tyramide was used as substrate. Double labelling of aCC, RP2 and dMP2 neurons was performed with anti-myc and Alexa 488 secondary antibody. For fluorescence quantification in aCC and RP2 motor neurons, heterozygous embryos and mutant siblings were isolated from the same embryo collection and identified on the basis of the presence (heterozygotes) or absence (homozygous mutants) of β-galactosidase in balancer chromosomes. Embryos were dissected and mounted on the same slide. Stacks of images were obtained as described above. Laser power and detector settings were optimized for detection of unsaturated fluorescent signal, and were kept constant for all the different genotypes in each experiment. Fluorescence within regions of interest was quantified with ImageJ.
Data are presented as mean values ± s.e.m. Using SPSS 16 software (SPSS), histograms were generated and statistical significance was determined. For analysis of genetic interaction between unc-5 and eve as well as unc-5 and grn, Kruskal–Wallis one-way analysis of variance was used. For all other comparisons, two-independent samples t-test was used.
unc-5 genomic fragments were amplified from genomic DNA and TA cloned into pCR8/GW/TOPO (Invitrogen) according to the manufacturer’s protocol. Fragments were then cloned using LR Clonase II Plus enzyme (Invitrogen) into pGWA-NLS-Venus. NLS-Venus was amplified and inserted (XbaI/XhoI) into a pUAS-attB vector with an inserted Gateway A cassette (PstI/BglII) to generate pGWA-NLS-Venus. All constructs were inserted by site-specific recombination into the attP2 site at 68A4 on 3L.
unc-5 and grn function in the same pathway to mediate motoneuron pathfinding
We have previously shown that the Unc-5 receptor is required in dMNs to reach its target muscles. In its absence, the ISN nerve will cross segment boundaries and even fasciculate with ISN from neighbouring segments as it presumably fails to sense the repulsive signal from the Netrins (Labrador et al., 2005), a phenotype similar to those observed for Netrin mutations (Mitchell et al., 1996). A homeobox transcription factor, Even-skipped (Eve) is largely responsible for the specification of aCC and RP2 dMNs (Doe et al., 1988; Landgraf et al., 1999; Fujioka et al., 2003) and regulates the expression of unc-5 (Labrador et al., 2005). In order to understand how guidance is transcriptionally regulated in individual motoneurons, we screened for other transcriptional regulators that presented a genetic interaction with unc-5. First, we determined whether unc-5 and eve were able to interact genetically in a transheterozygous combination. This is a genetic test in which the levels of two genes are halved and if the combination shows a phenotype not present on each single heterozygous, it strongly suggests that both genes are in the same pathway. Reducing unc-5 and eve levels to 50% did indeed result in a transheterozygous interaction (Fig. 1A,B; Table 1). This dose-sensitive interaction indicated that an unc-5 heterozygous background could be used to identify other unc-5 regulators. In particular, this system could be used to screen for different transcriptional regulators present in dMNs that control unc-5 during axon guidance.
One potential candidate is the GATA 2/3 homologue, Grn, because it is required for proper guidance of the aCC and RP2 dMNs to their muscles (Fig. 3D,E) (Garces and Thor, 2006). Furthermore, in grn mutants, motoneurons are properly specified because they are still able to induce Mad phosphorylation (Garces and Thor, 2006) and express several motoneuron markers (supplementary material Fig. S1). Thus, we generated transheterozygous combinations of unc-5 and grn. Although grn/+ or unc-5/+ seldom showed any guidance defect, transheterozygous unc-5/+; grn/+ embryos presented significant guidance defects in their ISN motoneurons (13%±2 s.e.m. P<0.001; Fig. 1C; Table 1), an average of 24% defective ISNs per embryo presenting a phenotype (Fig. 1C, Table 1). This result strongly suggests that Unc-5 and Grn function in the same pathway to mediate guidance of dMNs.
grn is required for unc-5 expression in aCC and RP2
This genetic interaction led us to propose that grn and unc-5 work together to mediate guidance of dMNs. These results also suggested that grn acts upstream of unc-5 and regulates its expression. To test this hypothesis, we analysed unc-5 mRNA levels in grn/+ and grn/grn mutant embryos. We determined unc-5 expression by in situ hybridization with fluorescent RNA probes and labelled dMN (aCC and RP2) cell bodies with protein markers (RN2Gal4 driving UASTauMycGFP) (Fujioka et al., 2003). This double-labelling technique allows us to determine levels of unc-5 mRNA and quantify its changes with single-cell resolution (Labrador et al., 2005). unc-5 is normally expressed in aCC and RP2 motoneurons in Drosophila embryos beginning at stage 12. Double labelling of dorsal-projecting motoneurons and unc-5 mRNA showed that by stage 14 unc-5 levels are substantially reduced in aCC and RP2 in grn/grn homozygous mutant embryos (Fig. 2B,D) when compared with their heterozygous siblings (Fig. 2A,C). unc-5 mRNA was reduced in grn mutant aCC neurons from 23.6± 4.8 to 10.7± 1.5 (P<0.02) and in grn mutant RP2 from 30.6±3.8 to 14.3±1.9 (P<0.005) (arbitrary fluorescence units ± s.e.m.) when compared with grn heterozygous (Fig. 2E). Importantly, unc-5 expression in grn homozygous mutants remained unchanged in cells in which grn is not normally expressed and is therefore not required for unc-5 expression, such as exit glia (Fig. 2F; supplementary material Fig. S2). These results strongly suggest that grn is required for unc-5 expression in both aCC and RP2 dMNs.
unc-5 partially rescues pathfinding defects in grn mutant embryos
unc-5 and grn are both required for proper guidance of dMNs although the grn mutant phenotype is more severe with >85% of the ISN nerves affected (Garces and Thor, 2006). The aCC and RP2 dMNs normally innervate muscles 1 and 2, respectively (Fig. 3A,B). However, in grn mutants the ISN nerve stops short of its most dorsal target and fails to innervate muscle 1 (Garces and Thor, 2006) (Fig. 3D,E). We reasoned that if Unc-5 functions downstream of Grn in dMNs and contributes to its pathfinding output, the grn guidance phenotype might be partially rescued by increasing unc-5 levels in aCC and RP2. Indeed, expression of HA-unc-5 in aCC and RP2 neurons in grn mutant embryos resulted in a partial rescue of the grn phenotype (Fig. 3F,G). We observed 10% more ISN branches projecting beyond muscle 2 and a reduction in defective segments from 83.6%±1.8 s.e.m. to 73%±3.1 s.e.m. (P<0.006) (Fig. 3G). These results further support the model that unc-5 is a downstream target of Grn, and is required for motor-guidance.
unc-5 expression is dependent on eve and grn
Both, eve and grn have been previously shown to be required for the proper specification of aCC and RN2. They can work in a linear pathway, as eve is required in aCC for grn expression, but they also work in independent pathways in both dMNs (Garces and Thor, 2006). If grn and eve are also working independently to induce unc-5 expression, we should be able to show this by analysing its expression in an eve/eve; grn/grn double mutant. We have previously shown that unc-5 levels are substantially reduced in eve mutant dMNs (Labrador et al., 2005); however, some cells still express unc-5 (11%; Fig. 4B,E asterisk). Therefore, we quantified the number of cells that still express unc-5 in the double-mutant embryos for eve and grn and were able to determine that the number of cells expressing unc-5 is further reduced (to 2%; Fig. 4C,F). This result suggests that eve and grn can also work in parallel pathways to induce unc-5 expression in dMNs.
An unc-5 embryonic neuronal enhancer drives expression in an eve-dependent manner
In order to identify the regulatory regions responsible for unc-5 expression in dMNs, we scanned the unc-5 locus from the stop of the preceding gene (Hr51) to the 5th intron within the unc-5 gene using different genomic fragments fused to the GFP or Gal4 genes (Fig. 5A). We identified a single region that could drive expression in neurons in the embryonic nervous system. This region drives expression in aCC, RP2 dMNs (Fig. 5B-D) and the segmental nerve a, (SNa) motoneurons (data not shown) in the same pattern as endogenous expression of unc-5 (Keleman and Dickson, 2001; Labrador et al., 2005).
We reasoned that if this region is responsible for unc-5 expression in dMNs, it should be regulated in the same way as the endogenous gene in those cells. To determine whether this enhancer is regulated by eve or grn, we stained for GFP signal driven by the enhancer in an eve- or a grn-null background. We showed that the enhancer is under the control of eve (Fig. 5E-H), as the GFP signal visible in eve/+ embryos in aCC and RP2 (Fig. 5E,F) is absent in eve/eve mutants (Fig. 5G,H). Importantly, the GFP signal in motoneurons from the segmental nerve (SN in Fig. 5F,H), where eve is not expressed nor required, is still present.
The neuronal enhancer drives expression of Gal4, which in turn drives expression of GFP. Given the stability of the Gal4 protein and its amplification effect on the UAS-driven reporter, we reasoned that we should be able to detect a difference only if the change is very considerable. We were, therefore, able to detect a difference in eve mutants as its elimination leads to an almost complete downregulation of endogenous unc-5 (Fig. 4) (Labrador et al., 2005). Not surprisingly, we failed to detect a significant effect in grn mutants as unc-5 downregulation in grn mutants is less severe than in eve/eve (Fig. 2). Alternatively, it is possible that another enhancer not identified in this study might be responsible for grn regulation of unc-5 in dMNs.
We can therefore conclude that we have identified a neuronal enhancer element responsible for the eve-dependent expression of unc-5 in dMNs.
Ectopic expression of grn or eve in dMP2 neurons is sufficient for unc-5 expression
During Drosophila embryonic nervous system development, grn is downstream of eve in aCC, although it works independently of eve in RP2 (Garces and Thor, 2006). Thus, there is a different requirement for eve and grn in each individual dMN. Our results show that grn is required in both neurons for unc-5 expression (Fig. 2), suggesting that grn might work sequentially or in a parallel pathway to eve to induce unc-5 transcription in a context-dependent manner. Additionally, our results also demonstrate that eve and grn also work independently to promote unc-5 expression. To test this hypothesis, we investigated whether grn or eve are sufficient to promote unc-5 expression in other neurons that would normally not express either of them (Fig. 6A, Fig. 7). We misexpressed grn or eve in dMP2 neurons, which are hindgut-innervating peptidergic motor neurons (Miguel-Aliaga and Thor, 2004), with the dMP2Gal4 driver and examined unc-5 expression by in situ hybridization. Upon expression of either grn or eve in dMP2 neurons, we observed unc-5 mRNA expression in 74% and 62% of the neurons, respectively (Fig. 6B,D). These results show that grn is sufficient to induce unc-5 expression in vivo in a similar way as eve. Interestingly, we failed to detect grn in dMP2 neurons when eve was misexpressed or eve when grn was misexpressed (Fig. 7), suggesting that eve or grn can induce unc-5 expression independently of each other. When both transcription factors were expressed in dMP2 neurons there was a further increase in the level and number of neurons expressing unc-5 (86% of neurons; Fig. 6C), further supporting the hypothesis that in dMP2 neurons both transcription factors can work in parallel to induce unc-5 transcription and the coordinated action of both leads to a more robust induction of unc-5.
A transcription code for axonal exit towards the muscle field is dependent on unc-5
Changing the transcription factor code in dMP2 neurons by expressing grn and eve together forces their axons to leave towards the muscle field (Garces and Thor, 2006) and follow the ISN nerve similarly to dMN projections (Fig. 8C) but each factor alone does not have this effect (Fig. 8B) (Garces and Thor, 2006). Likewise, unc-5 misexpression in dMP2 neurons can also lead to aberrant axonal exit (Fig. 8D), similarly to apterous interneurons (Keleman and Dickson, 2001; Allan et al., 2003). Because eve and grn are upstream of unc-5 in dMNs and unc-5 is required as part of their guidance output (Figs 1, 2), the guidance behaviour of these transcriptionally ‘reprogrammed’ neurons might also be dependent on the Unc-5 receptor repulsive output. To test this hypothesis, we misexpressed both eve and grn in dMP2 neurons and reduced unc-5 levels or completely eliminated unc-5 (Fig. 8E,F). Aberrant axon exit was significantly reduced (P<0.02) from 34%±5.1 to 21%±2.4 s.e.m in an unc-5 heterozygous background and further reduced to a 6%±2.1 s.e.m exit (P<0.0001) in an unc-5 null background (Fig. 8E,F), indicating that unc-5 is required for projection of dMP2 neurons into the muscle field. These data, together with the ability of each individual transcription factor to induce unc-5 (Fig. 6), suggests that unc-5 is a common determinant of guidance regulated by eve and grn. Furthermore, the dose-dependent effect of unc-5 in the suppression of the exit phenotype suggests that levels of the receptor are crucial to promote exit and only the combined effect of both transcription factors results in high enough levels of Unc-5.
grn controls motor axon guidance through the Unc-5 receptor
Different transcriptional codes regulate axon guidance but the guidance systems they regulate are still unknown. The GATA factor, Grn, is a major determinant of guidance within dorsal MNs. Garces and Thor have previously determined a strong guidance phenotype in grn mutants, in which ISN axons fail to reach their targets in >85% of the segments (Garces and Thor, 2006). However, none of the downstream molecules required for guidance downstream of grn has been identified to date (Landgraf and Thor, 2006). Several lines of evidence indicate that the Unc-5 receptor mediates guidance downstream of the GATA transcription factor Grn. First, both genes interact genetically in trans and this type of genetic interaction is often seen between two genes the gene products of which directly interact, such as Slit and Robo (Kidd et al., 1999). Second, there is also a partial requirement of grn for unc-5 expression in dMNs as unc-5 mRNA levels are reduced in aCC and RP2 in grn mutants. Further support for the role of unc-5 downstream of grn comes from the partial rescue of the grn phenotype obtained by exogenously providing unc-5 specifically in aCC and RP2.
eve mediates dMN guidance through unc-5
Among transcriptional codes that regulate motor-axon pathfinding, specific Lim-HD codes are required for the proper guidance of vertebrate motoneurons, in part, through the regulation of the EphA (Kania and Jessell, 2003), EphB (Luria et al., 2008) or FGF receptors (Shirasaki et al., 2006). In Drosophila, Nkx6 (HGTX – FlyBase) is important for vMN specification and has been proposed to promote guidance through the expression of fasciclinIII (Fasciclin 3 – FlyBase) (Broihier et al., 2004). Similarly, eve regulates guidance of dMNs through unc-5 (Labrador et al., 2005). In at least some of these situations the regulation is unlikely to be direct because the regulators are thought to mediate their function through a repressive activity (Muhr et al., 2001; Fujioka et al., 2003). Indeed, eve repressive activity has been shown to be responsible for its guidance function in aCC and RP2 (Fujioka et al., 2003), suggesting that its function might be to repress the expression of other transcription factors, such as HB9 (Exex – FlyBase) (Fujioka et al., 2003; Broihier et al., 2004), which would confer those motoneurons with a ventral fate. Our analysis of the unc-5 neuronal enhancer regulated by eve failed to identify any conserved Eve consensus homeodomain binding site, suggesting that eve regulation of unc-5 is not mediated through a direct binding to this element.
eve and grn can regulate unc-5 through independent parallel pathways
Combinatorial codes of transcription factors play an instructive role in the generation of subclass diversity within the vertebrate spinal cord (Dasen and Jessell, 2009; Bonanomi and Pfaff, 2010). In invertebrates, in which it is possible to analyse individual motoneurons within a subclass, a further level of complexity is revealed. Within the subclass of motoneurons that project to dorsal muscles, aCC and RP2, subclass determinants (eve, grn and zfh1) can work in a sequential order in aCC specification or independently, in parallel pathways, within RP2 (Garces and Thor, 2006). Whereas Zfh1 is a general factor required in all motoneurons (Layden et al., 2006), eve and grn are specific to dMNs. It is plausible that each one of them might be important for specific aspects of specification within the same subclass of neurons (Garces and Thor, 2006) but together might be responsible for the regulation of common targets such as unc-5. Although grn is required in both aCC and RP2 for unc-5 expression, it is not the only factor required because unc-5 mRNA is not completely absent from those cells and it also requires the presence of eve. In fact, grn or eve can independently induce unc-5 expression in dMP2 neurons but only both factors expressed in combination are able to induce axon exit towards the muscle field. This combinatorial expression of eve and grn might bring unc-5 above the threshold required for exit. unc-5 levels are definitely important for dMP2 exit because removing 50% of the gene dosage significantly suppresses the exit phenotype and this suppression is almost complete in an unc-5 null background. Transheterozygous interactions identified between unc-5 and eve, or unc-5 and grn also suggest that their levels are tightly controlled. As both eve and grn can regulate unc-5 it is likely that their combined activity in aCC and RP2 is essential to express the required levels of unc-5 in both neurons. A model of unc-5 regulation by grn and eve in each individual neuron is shown in Fig. 9.
The combinatorial nature of substrate recognition during motoneuron guidance and targeting is well established (Winberg et al., 1998; Kurusu et al., 2008). Mutants for guidance receptors or ligands often show partially penetrant phenotypes (Vactor et al., 1993) and this includes the unc-5 phenotype in dMNs (Labrador et al., 2005). By contrast, phenotypes observed for transcription factor mutants that affect dMN guidance are far more severe. For example, Lim1 regulates EphA4 in vertebrate motoneurons but the EphA4 mutant phenotype is less severe than Lim1 mutant phenotype (Kania and Jessell, 2003). Similarly, 100% of ISNs in eve mutants (Landgraf et al., 1999; Fujioka et al., 2003) or 85% in grn mutants (Garces and Thor, 2006) are affected, suggesting that they regulate several guidance systems. It will be interesting to investigate the full array of guidance systems regulated by the eve-grn code. One potential candidate is FasII (Fas2 – FlyBase) (Landgraf et al., 1999) as it is an essential molecule for the pioneer function of aCC and RP2 (Sánchez-Soriano and Prokop, 2005).
In summary, we have identified the guidance receptor Unc-5 as a novel target of the GATA transcription factor Grn. unc-5 is key common target that mediates guidance downstream of both eve and grn. Furthermore, both transcription factors can promote transcription of unc-5 independently, suggesting that their combined action is essential to attain the proper expression levels of the Unc-5 receptor in dMNs. Future investigations in cell type-specific expression profiling and chromatin immunoprecipitation sequencing (ChIP-seq) analysis will hopefully unravel other guidance systems regulated by the eve-grn code in dMNs. This knowledge will bring us closer to understanding the cell-specific transcriptional regulation of guidance.
We would like to thank Miki Fujioka, Stefan Thor, Alain Garces, Chris Doe, James Skeath, Castelli-Gair Hombria, Greg Bashaw, Brian McCabe and Zhenghe John Wang for fly stocks and reagents. We are grateful to Kevin Mitchell, Mario Fares, Mani Ramaswami and members of the lab for critical reading of the manuscript.
This work was supported by a Principal Investigator Grant [07/IN.1/B913] and the Research Frontiers Program [08/RFP/NSC1617] from Science Foundation Ireland [to J.-P.L.]; an IRCSET Embark Initiative Postgraduate Scholarship [to A.A.-Z.]; and an IRCSET postdoctoral fellowship [to J.H.].
Competing interests statement
The authors declare no competing financial interests.