Synaptic transmission in neurons that express the Drosophilaatypical soluble guanylyl cyclases, Gyc-89Da and Gyc-89Db, is necessary for the successful completion of larval and adult ecdysis

The intracellular messenger cyclic 3′,5′ guanosine monophosphate (cGMP) regulates a wide variety of physiological processes in vertebrates and invertebrates (Lucas et al., 2000; Morton and Hudson,2002). The synthesis of cGMP is catalyzed by guanylyl cyclases that can be classified as either receptor guanylyl cyclases, which are integral membrane proteins, or soluble guanylyl cyclases, which are localized to the cytoplasm (Lucas et al.,2000). Soluble guanylyl cyclases can be further subdivided into conventional soluble guanylyl cyclases, which are the targets of the gaseous messenger nitric oxide (NO) (Lucas et al.,2000), and the atypical soluble guanylyl cyclases that are poorly activated by NO (Morton,2004a) and have been proposed to act as oxygen sensors(Gray et al., 2004). The Drosophila genome contains five genes that code for subunits of soluble guanylyl cyclases (sGCs) (Morton,2004a). Two of these, Gycα-99B and Gycβ-100B, form the conventional NO-sensitive sGC and the other three genes code for atypical subunits that form oxygen-sensitive sGCs (Morton, 2004b). We have previously shown that the atypical sGC subunits, Gyc-88E, Gyc-89Daand Gyc-89Db, are expressed in a subset of central and chemosensory peripheral neurons (Langlais et al.,2004; Morton et al.,2005a). Based on the expression pattern and biochemical properties of these subunits, we predicted that the sGCs formed from the atypical subunits would be involved in olfaction, gustation and/or oxygen sensation(Vermehren et al., 2006).

In this study we began to examine the function of the neurons that express the atypical sGCs Gyc-89Da and Gyc-89Db. Using the predicted promoter regions of these genes we generated flies that express the yeast transcription factor GAL4 (Brand and Perrimon, 1993)under the control of the promoters for each GC subunit. We then used these flies to express the light chain of tetanus toxin (TNT) under the control of an upstream activation sequence (UAS) that is regulated by GAL4 to block synaptic transmission in the cells that express Gyc-89Da and Gyc-89Db. The results show that at least some of these neurons are required for the successful initiation and completion of larval and adult ecdysis.

A variety of GAL4 and UAS fly lines were kindly provided by the following labs: Ap-GAL4 (apterous-GAL4) from Paul Taghert, Washington University, St Louis, MO, USA; CCAP-GAL4 (CCAP promoter) from John Ewer, Cornell University,NY, USA; dimm-GAL4 (dimmed) from Randy Hewes, University of Oklahoma, Norman,OK, USA; EHups-GAL4 (eclosion hormone promoter driving GAL4) from Sue McNabb,University of Washington, Seattle, WA, USA; UAS-TNTa and UAS-TNTi lines (lines 1171 and 1178 that express the active and inactive forms of the light chain of tetanus toxin under UAS control) were provided by Cahir O'Kane and Sean Sweeney, University of Cambridge, UK; and UAS-PDE5 line (bovine phosphodiesterase 5) from Shireen Davies, University of Glasgow, UK. The UAS-dsRed, tubp-GAL80ts lines were obtained from the Bloomington Stock Center. All crosses were carried out using standard methods at 25°C(Greenspan, 1997) unless otherwise indicated.

The Gyc-89Da promoter (p89Da) lines were generated by cloning the 1.9 kb sequence upstream of the translational start site of Gyc-89Dathat included the 5′ UTR and first intron. The 1931 bp PCR fragment was amplified from genomic DNA using the primers 5′-GGGTTCTTTTTTGTAGAGAAA-3′ and 5′-GGTTTATGTTTCTGAAAAAATG-3′ and Taq DNA polymerase(Invitrogen, Carlsbad, CA, USA). The PCR fragment was gel purified and subcloned directly into the Topo-TA vector (Invitrogen). The Gyc-89Dapromoter fragment was then excised with BamHI and NotI restriction enzymes (Invitrogen) and subcloned into the P-element vectors pPTGAL (Sharma et al., 2002)and green fluorescent protein (GFP) pGreen Pelican(Barolo et al., 2000). The Gyc-89Db promoter (p89Db) lines were generated in a similar manner by cloning the 3398 bp sequence 5′ of Gyc-89Db, using the primers 5′-GCACCTGTGGCTCTCTTA-3′ and 5′-GATGGGGCAGGATGTGAA-3′. The PCR fragment was then subcloned into the pPTGAL and GFP vectors. DNA for microinjection was prepared by ultracentrifugation using a CsCl gradient. About 200–400 wy¹;; embryos were co-injected with 0.2 μgμl^–1 of helper plasmid (ρπ25.7 Δ2–3 wc)and 1 μg μl^–1 of either the GFP or GAL4 constructs in 0.85 mmol l^–1 NaH₂PO₄, 9.15 mmol l^–1 Na₂HPO₄, 5 mmol l^–1 KCl buffer, and maintained at 18°C until eclosion. Eleven insertion lines were recovered for p89Da-GAL4, two for p89Da-GFP, four for p89Db-GAL4 and two for p89Db-GFP. Chromosomal locations of the P-element insertions were determined by standard crossing methods to a ubiquitous balancer line(w;CyO/Kr^If-1;Tm6B,Tb⁺/D¹).

In situ hybridization of embryos using digoxygenin-labeled probes was carried out as previously described(Langlais et al., 2004).

Tissue or whole larvae were mounted in aqueous mounting media (gel mount,Sigma, St Louis, MO, USA) and viewed on a Nikon Eclipse E800 fluorescence microscope equipped with argon and helium/neon lasers (Radiance 2100, Bio-Rad,Hercules, CA, USA) to excite the GFP and dsRed proteins (emission filters 500–530 nm and 570–650 nm, respectively). Usually, a stack of 15 sections was taken (z-axis step size, 2.5 μm) using Laser Sharp 2000 software (Bio-Rad) and image processing was carried out using ImageJ(http://rsb.info.nih.gov/ij/). Identification of neurons expressing GFP was made based on a combination of their morphologies and their relative position.

The mortality of developing animals expressing the active or inactive forms of TNT was determined as follows. To determine the hatching success rate,flies were allowed to lay eggs on standard fruit plates(Sullivan et al., 2000)overnight at 25°C. Embryos were collected and arranged in rows on a new fruit plate and allowed to hatch for 36 h at 25°C. Hatched and un-hatched embryos were then counted. To examine the survival of first instar larvae to the end of the third instar stage and into the pupal stage, a known number of first instar larvae were placed in a standard food vial or on a custom-made food disc in a plastic 8.5 cm Petri dish, with a few granules of dried yeast. The food disc was prepared by melting standard fly food, pouring it onto a large glass plate and compressing it between a second glass plate separated by 3 mm spacers. When solidified, 5.6 cm diameter discs were then cut from this sheet. Larvae placed in vials or on discs were kept at 25°C until pupariation and counted. The developmental stage and location of any dead larvae were recorded. To examine defects during eclosion, developing adults were collected from vials at stage P15 and lined up on a strip of double-sided clear tape on a glass slide. Just prior to eclosion, the operculum was manually removed from each pupa, and the pre-eclosion staging markers(Kimura and Truman, 1990; McNabb et al., 1997) were observed through a dissecting microscope. At least 15 experimental and control animals from three separate crosses were examined for pre-eclosion defects.

The temperature-sensitive inhibitor of GAL4, GAL80ts(Zeidler et al., 2004), was used in some experiments to control temporal expression of TNT. Using standard crossing schemes (Greenspan,1997), a homozygous fly line was created that contained GAL80ts driven by the tubulin promoter (tubP-GAL80ts) on chromosome 2 and the p89Da-GAL4 insertion on chromosome 3. These flies were then crossed with UAS-TNTa, UAS-TNTi or UAS-dsRed flies, resulting in offspring that contained tubP-GAL4ts, p89Da-GAL4 and UAS-TNTa, UAS-TNTi or UAS-dsRed. These crosses were carried out at 18°C, the permissive temperature for GAL80ts. At 18°C, GAL80ts inhibits the action of GAL4. Experimental vials, as described in Results, were switched at specific stages to an incubator set at 30°C, the non-permissive temperature for GAL80ts(Zeidler et al., 2004) that allows GAL4 to function and activate transcription of UAS-controlled transgenes.

Our previous studies using whole-mount in situ hybridization labeling in embryos showed that all three atypical sGC subunit genes, Gyc-89Da, Gyc-89Db and Gyc-88E, were expressed in a subset of peripheral and central neurons(Langlais et al., 2004; Morton et al., 2005a). By examining the number and position of these cells we concluded that these three genes were co-expressed in many of the peripheral sensory neurons(Langlais et al., 2004; Morton et al., 2005a). To better define these co-expression patterns, we used upstream regions of each gene to drive expression of GFP in transgenic flies (designated p89Da-GFP and p89Db-GFP). The portions of genomic DNA 5′ to Gyc-89Da and Gyc-89Db (1.9 and 3.4 kb,respectively) were sufficient to drive GFP expression in embryos in a pattern that closely resembled that seen with in situ hybridization(Fig. 1). By contrast, however,a 5 kb length of genomic DNA 5′ to Gyc-88E was insufficient to drive expression of GFP in any cells. Peripheral neurons that expressed GFP in p89Da-GFP and p89Db-GFP larvae included a small subset of cells in the dorsal and terminal ganglia (Fig. 1G–I), which innervate the olfactory and gustatory organs,respectively (Stocker, 1994; Heimbeck et al., 1999),neurons innervating external sensilla along the lateral body wall(Fig. 1D–F) and neurons that innervate the terminal sensory organs(Fig. 1J–L)(Langlais et al., 2004; Morton et al., 2005a). These expression patterns closely matched those seen using in situhybridization in late embryos (Langlais et al., 2004) (Fig. 1).

Most of the peripheral neurons that expressed GFP under the control of the Gyc-89Da and Gyc-89Db promoters had stereotyped positions and the relatively low number of cells that expressed GFP made it possible to compare these preparations with the in situ hybridizations carried out in embryos. A comparison of the number of cells in each group labeled by each method is shown in Table 1. In general, there was a very good match for the number of cells in each group between the embryo in situ hybridization data and direct observation of GFP fluorescence driven by each of the sGC promoters. In the dorsal and terminal ganglia a small number of neurons (five or less) were labeled with both in situ hybridization and direct visualization of GFP for both Gyc-89Da and Gyc-89Db(Table 1). Sensory neurons innervating external sensilla along the larval body wall can be broadly divided into three groups – dorsal (des cells), lateral (les cells) and ventral (ves cells) (Bodmer et al.,1987). In our previous studies using a combination of in situ hybridization and immunocytochemistry with a neuronal marker in embryos, we showed that Gyc-89Da and Gyc-89Db were expressed in a single les and a single ves cell in the thoracic segments and Gyc-89Db was expressed in a tracheal dendritic neuron (td cell) in the first two abdominal segments only(Langlais et al., 2004; Morton et al., 2005a). Direct visualization of GFP expression driven by the two promoters also labeled these cells. Two cells on each side of the larvae were observed in each of the thoracic segments of p89Da-GFP and p89Db-GFP larvae(Table 1). One of these cells was located in a more lateral position and the other in a more ventral position corresponding to the positions of the les and ves cells,respectively. The td neuron that was identified in the first two abdominal segments of embryos using in situ hybridization of Gyc-89Db(Langlais et al., 2004) was also observed in p89Db-GFP larvae with the exception that these cells were also seen in abdominal segment 3(Table 1). In the terminal abdominal segments there are seven sensory organs on either side in the larvae(Stocker, 1994), and in situ hybridizations in embryos showed that a single neuron innervating six of these structures expressed Gyc-89Da and Gyc-89Db(Langlais et al., 2004; Morton et al., 2005a). Direct visualization of GFP in p89Da-GFP and p89Db-GFP larvae also revealed these neurons, and we were able to detect a single neuron in each of the seven sensory organs (Table 1). These experiments show that there is a very good match between expression of Gyc-89Da and Gyc-89Db determined by in situhybridization and expression of GFP driven by portions of 5′ genomic DNA.

Our previous studies (Langlais et al.,2004; Morton et al.,2005a) used cell counts to determine the overlap in the expression patterns of the three atypical sGC subunits. These experiments indicated that in the majority of peripheral cells Gyc-89Da and Gyc-89Dbwere co-expressed with Gyc-88E and suggested that in many cells all three subunits were co-expressed (Langlais et al., 2004; Morton et al.,2005a). To determine the co-expression patterns of Gyc-89Da and Gyc-89Db more accurately, we generated transgenic flies that expressed the yeast transcription factor GAL4 under the control of the upstream regions of these genes (designated p89Da-GAL4and p89Db-GAL4) and crossed them with flies that expressed red fluorescent protein (dsRed) under UAS control. Examination of larvae from these crosses showed the same number of peripheral neurons as seen with the p89Da-GFP and p89Db-GFP lines, respectively(Table 1). We then generated larvae that contained three transgenes: p89Da-GFP, p89Db-GAL4 and UAS-dsRed, and examined the neurons under both green and red fluorescence. In larvae expressing p89Da-GFP, p89Db-GAL4 and UAS-dsRed we consistently observed four neurons that expressed Gyc-89Da and four neurons that expressed Gyc-89Db in the dorsal ganglion, but none of these neurons co-expressed the two subunit genes(Fig. 2A). Similarly, in the terminal ganglion, three to four neurons expressed Gyc-89Da and three to four neurons expressed Gyc-89Db, but we never observed neurons in the terminal ganglion that expressed both subunits(Fig. 2A). In addition to the neurons in the dorsal and terminal ganglia within the larval head, we consistently observed co-expression in two neurons that appeared to be part of the monoscolopidial chordotonal organs(Fig. 2A), which lie adjacent to the terminal ganglia (Campos-Ortega and Hartenstein, 1997). Expression of Gyc-89Db alone was also seen in two to three multidendritic arborization neurons(Grueber et al., 2002) in the head region.

In contrast to the terminal and dorsal ganglia, co-expression of Gyc-89Da and Gyc-89Db was seen in peripheral neurons in the thoracic segments. Both cells in all three thoracic segments co-expressed Gyc-89Da and Gyc-89Db(Fig. 2B, Table 1). The td neuron in the abdominal segments, however, only expressed Gyc-89Db and did not show Gyc-89Da expression (Fig. 2B, Table 1). In the caudal segments, Gyc-89Da and Gyc-89Dbwere co-expressed in neurons that innervate peg sensilla on the sensory cones(Fig. 2C). There are seven sensory cones on each side of the larvae, which have been described as possible chemosensory detectors (Stocker,1994). Each sensory cone appears to be innervated by one to two neurons and only a single neuron in each cone co-expressed the two subunit genes. A neurite was clearly visible extending to the peg sensilla at the tip of each cone (Fig. 2C,inset).

Gyc-89Da and Gyc-89Db were also expressed in the larval CNS, but no co-expression of the two subunits was detected in central neurons(Fig. 2D). Numerous cells were labeled in the brain lobes and ventral nerve cord, with a larger number of cells that expressed Gyc-89Db compared with Gyc-89Da observed. The only cells in which Gyc-89Da and Gyc-89Db were co-localized were in the ring gland (neurosecretory organ), where several intrinsic secretory cells of the corpora cardiaca expressed both subunits (data not shown).

To analyze the functions of the neurons that express Gyc-89Da and Gyc-89Db we blocked synaptic transmission in these cells by expressing the light chain of TNT under UAS control(Sweeney et al., 1995). TNT abolishes transmitter release by cleaving the synaptic vesicle protein synaptobrevin (Schiavo et al.,1992; Link et al.,1992). Identification of behaviors that are disrupted when TNT is expressed in subsets of neurons in Drosophila provides a simple tool for examining the function of these neurons(Sweeney et al., 1995). When we expressed the active form of TNT (TNTa) in the neurons that express Gyc-89Da and Gyc-89Db the animals failed to develop into adult flies and died at three specific developmental stages associated with the molt (Fig. 3). As a control we also used a fly line that expressed an inactive form of TNT (TNTi) under UAS control (Sweeney et al.,1995). About 90% of eggs generated from these crosses survived to adult (Fig. 3). For those neurons that expressed Gyc-89Da, expression of TNT caused slightly reduced hatching rates (84.9±1.8% compared with 92.3±0.4% for expression of TNTi, P<0.05). The majority of the hatchlings survived to the end of the third instar and about 50% died at the beginning of metamorphosis, at about the time of pupariation. All of the animals that successfully pupated died just prior to adult eclosion (emergence from the pupal cuticle). By contrast, expression of TNT in the Gyc-89Db-expressing neurons caused all the animals to die at the first–second larval transition (Fig. 3).

To determine whether there were general behavioral deficits during larval stages, we tested the third instar p89Da-GAL4 × UAS-TNTa larvae for their general locomotion on 1% agarose plates(Osborne et al., 1997) and observed no significant differences between those expressing TNTa and those expressing TNTi in distance traveled (2.37±0.26 and 1.99±0.16 cm min^–1, respectively: N=9 for each genotype,Student's t-test P=0.2372). Although locomotion and growth appeared normal, we found substantial numbers of late third instar larvae that expressed TNTa in the Gyc-89Da neurons deeply buried in the food, whereas most of those that expressed TNTi had all crawled up the wall of the vial and formed puparia. We previously suggested that the Drosophila atypical sGCs act as neuronal O₂ sensors because they are activated by reduced O₂ concentrations(Morton, 2004b). In addition,a related sGC in C. elegans (GCY-35) has been shown to mediate O₂-sensitive behaviors (Gray et al., 2004). This suggested that larvae with inactive Gyc-89Da neurons might not be able to sense that they were deep in the food and suffocating. To prevent the TNTa-expressing animals from burying deeply into the food, Petri dishes containing a thin layer of food (∼3 mm thick) were used instead of the normal food vials (3 cm depth of food). Under these conditions, no significant differences in the number of animals that successfully pupated were observed between those expressing TNTa and those expressing TNTi (Fig. 4).

Almost all (98.2±0.8%) of the p89Da-GAL4 × UAS-TNTa larvae that successfully formed puparia failed to eclose as adult flies regardless of whether they were raised on thin discs of food or in normal vials. To investigate more closely the causes of the failure to eclose, we examined animals during the final few hours before eclosion for the appearance of developmental markers that precede eclosion(Kimura and Truman, 1990). Animals were examined under a dissecting microscope and the opercula removed to facilitate detection of the markers. About 8 h prior to eclosion the head of the developing adult is smooth, indicating that molting fluid is still present between the pupal and adult cuticle. About 3 h prior to eclosion the head has a grainy appearance, which indicates that molting fluid absorption is complete. The tracheal system and the space between the adult and pupal cuticles fill with air about 1 h before eclosion leading to a whitish appearance of the head. This is followed at about 40 min before eclosion by the extension of the ptilinum from the front of the head, which ruptures the thin pupal cuticle. All of these stages occurred in the proper order in p89Da-GAL4 × UAS-TNTa and p89Da-GAL4 × UAS-TNTi animals. Some animals were also removed from the pupal cases at the onset of tracheal air filling and examined under a microscope to determine the extent of tracheal air filling throughout the body, which appeared normal and was comparable to that of similarly staged control animals.

About 40 min after ptilinum extension, the p89Da-GAL4 × UAS-TNTi control animals began inflating their heads and pushed forward, which enabled the animals to rupture and push open the operculum. After several strong head inflations (10–20 s), the p89Da-GAL4 × UAS-TNTi control animals began the stereotyped abdominal, posterior-directed peristalsis movements(eclosion behavior) leading to complete emergence from the puparium. The eclosion of these control animals is comparable to that of wild-type Canton-S animals (McNabb et al., 1997). The animals that expressed the active form of TNT (p89Da-GAL4 × UAS-TNTa) also began head inflations about 40 min after ptilinum extension, but the magnitude of the inflations appeared much weaker. These weaker head inflations rarely succeeded in completely opening the operculum,when it was left intact. Several head inflations/deflations occurred and the animal did push forward noticeably, but the abdominal peristalsis component of the eclosion motor program was never initiated in these animals. Interestingly, head inflations/deflations occurred for up to 4 more hours. During this time, starting about an hour after the first head inflations, the antennae occasionally twitched, and the abdomen sporadically contracted several times a minute, but these movements did not resemble the coordinated series of peristaltic contractions seen in control animals. Even when the operculum was removed just prior to the expected time of eclosion, the animals expressing TNTa in the Gyc-89Da neurons never succeeded in escaping from the puparium.

After wild-type flies have eclosed they undergo wing inflation and cuticle tanning (see Ewer and Reynolds,2002). To determine whether animals that expressed TNTa in the Gyc-89Da neurons completed these processes in the absence of eclosion, we removed them from their puparia and observed them for up to 2 h. These animals remained immobile, never showed any wing inflation and after 2 h still had un-tanned cuticle. By contrast, animals that expressed TNTi inflated their wings and tanned their cuticles within 30 min of eclosion.

In contrast to the p89Da-GAL4 × UAS-TNTa progeny, the majority of which survived to the third instar, when TNTa was expressed in the cells that express Gyc-89Db, all the larvae died before they reached the second instar (Fig. 3). Fewer of the p89Db-GAL4 × UAS-TNTa progeny hatched compared with controls expressing TNTi (85.3±1.7% vs 92.6±1.3%, Student's t-test, P<0.05) and the newly hatched larvae expressing TNTa appeared to become increasingly sluggish after a day of feeding on the surface of the food. Although the larvae appeared morphologically normal they all failed to reach the second larval instar. About 80% of the p89Db-GAL4× UAS-TNTa larvae that hatched survived through the feeding phase and began the developmental events that lead to ecdysis (83.2±1.4% compared with 97.4±0.3% TNTi controls, Student's t-test, P<0.05). Of the 16.8% of larvae that died during the feeding phase, most were found buried under the food (63% buried vs 37% on top). This appeared to be similar to the third instar larvae expressing TNTa in the Gyc-89Da neurons, many of which appeared to drown in their food,possibly because they failed to detect that they had become anoxic.

There are several morphological markers that have been described that precede larval ecdysis in Drosophila(Park et al., 2002; Clark et al., 2004). About 1 h prior to the first larval ecdysis the new mouth parts of the second instar larvae become visible, a stage referred to as double vertical plates [dVP(Park et al., 2002)]. This is followed about 10 min later by the collapse of the old tracheal lining and almost immediately by the trachea filling with air. Shortly after tracheal air filling the pre-ecdysis and ecdysis behaviors commence(Park et al., 2002). Over 99%of the p89Db-GAL4 × UAS-TNTa larvae that survived past the first instar feeding phase were found with transparent fluid-filled double trachea and dVP. No animals were found with air-filled tracheae or that initiated pre-ecdysis or ecdysis peristaltic movements. We presume that the larvae found with dVP and transparent, fluid-filled trachea probably died from suffocation as they were unable to air fill their tracheae.

The endocrinology underlying the initiation of ecdysis and eclosion in Drosophila has been extensively studied and is known to be regulated by the release and action of at least three peptide hormones: ecdysis triggering hormone (ETH), eclosion hormone (EH) and crustacean cardiac acceleratory peptide (CCAP) (McNabb et al., 1997; Park et al.,2002; Clark et al.,2004). Deletion of the gene encoding these peptides or elimination of the cells that secrete the peptide lead to similar phenotypes to those described above for expression of TNTa in the cells that express Gyc-89Da and Gyc-89Db. For example, deletion of the ETH gene leads to lethality at the first larval ecdysis with most of the animals dying after tracheal inflation and some animals showing some`ecdysis-like' movements, although none initiated pre-ecdysis movements(Park et al., 2002). Killing the EH cells leads to larval death, with most of the larvae dying with fluid-filled tracheae, although the timing of larval pre-ecdysis and ecdysis behaviors appeared unaffected (McNabb et al., 1997; Clark et al.,2004). Furthermore, EH cell knockouts led to the disruption of adult eclosion, notably with a longer period between ptilinum extension and eclosion (McNabb et al.,1997). It is possible that the effects of TNTa expression in the Gyc-89Da and Gyc-89Db neurons were due to a failure of the release of some or all of these peptides. The simplest explanation for this would be if TNTa was expressed in the peptidergic cells directly preventing peptide release. ETH is expressed in the epitracheal gland located along the major tracheal branches(O'Brien and Taghert, 1998; Park et al., 2002). No expression of GFP or dsRed was seen in these cells but expression was seen in the brain and ventral nerve cord in locations that might correspond to the EH and/or CCAP cells (Fig. 2)(Langlais et al., 2004).

To determine whether there was overlap in the expression of EH or CCAP with either Gyc-89Da or Gyc-89Db we generated larvae that contained p89Da-GFP or p89Db-GFP with UAS-dsRed and the upstream region of EH coupled to GAL4(McNabb et al., 1997) or the upstream region of CCAP coupled to GAL4(Park et al., 2003). Fig. 5 shows images of the CNS from these larvae. EH is expressed in a single pair of cells in the ventral midline of the brain (Horodyski et al.,1993). Fig. 5A,Bclearly shows that there was no overlap in the expression of dsRed driven by the upstream region of EH with GFP driven by the promoter of either Gyc-89Da or Gyc-89Db. CCAP is expressed in a row of paired cells along the lateral edges of the ventral nerve cord(Park et al., 2003) and again there was no overlap with the expression of Gyc-89Da or Gyc-89Db (Fig. 5C,D).

The row of cells lying adjacent to the CCAP cells appears to be in a similar position to peptidergic neurons that express the transcription factor`dimmed' (Hewes et al., 2003). To determine whether any of the Gyc-89Da- or Gyc-89Db-expressing cells also expressed dimmed, we generated larvae that contained the dimmed-GAL4, UAS-dsRed and p89Da-GFP or p89Db-GFP transgenes. Examination of the nervous tissue of these larvae showed that paired cells in the ventral nerve cord co-expressed Gyc-89Da and dimmed (Fig. 5E). These cells appeared to be in a similar position to the dorsal furin-containing neurons (d1–d11) described by Hewes et al.(Hewes et al., 2003). These furin-expressing cells also express the transcription factor `apterous' and it has been suggested that neuropeptides from these cells are involved in the control of larval ecdysis (Park et al.,2004). We also tested whether there was overlap in the expression patterns of the Gyc-89Da or Gyc-89Db cells using an apterous enhancer trap line (Park et al., 2004). Examination of the larval CNS from animals bearing UAS-dsRed, ap-GAL4and p89Da-GFP or p89Db-GAL4 showed no overlap in expression of either Gyc-89Da or Gyc-89Db and apterous (data not shown). It therefore appears that the subset of Gyc-89Da neurons that are dimmed positive are apterous negative.

Several studies have shown that cGMP mediates the action of EH and although the identity of the guanylyl cyclase that is activated is not known, there is evidence that it is an atypical sGC (see Morton and Simpson, 2002). Because we were specifically targeting neurons that expressed this subclass of sGC, it was possible that cGMP in these neurons was required for, or modulated, ecdysis and eclosion. To test for this we expressed a cGMP-specific phosphodiesterase (bovine PDE5) in the Gyc-89Da and Gyc-89Db neurons using a UAS-bPDE5 fly line (Broderick et al.,2004). Progeny of these crosses all developed and eclosed normally with no differences in either the time taken for development or the number of animals that successfully eclosed compared with either of the parental lines(data not shown). We also closely observed the final stages of adult eclosion as described above and could detect no differences in the time taken for any of the stages prior to eclosion (data not shown).

The action of TNTa in the Gyc-89Da neurons to prevent eclosion could be due to its acute action at the time of eclosion or it could be due to disruption of an earlier developmental event that is subsequently required for eclosion. To distinguish between these two possibilities we co-expressed GAL80ts, a temperature-sensitive inhibitor of GAL4, with p89Da-GAL4 and UAS-TNTa to control the temporal expression of TNTa. The yeast GAL80 protein blocks GAL4 activity by binding to its transcriptional activation domain, a function that is retained when it is expressed in Drosophila (Yocum and Johnston,1984; Zeidler et al.,2004). A temperature-sensitive version of GAL80 (GAL80ts) was created by inserting a temperature-controlled yeast intein into a crucial region of the GAL80 peptide and placing this construct under the control of a ubiquitous tubulin promoter (tubP-GAL80ts)(McGuire et al., 2003). We then generated flies that were homozygous for p89Da-GAL4 and tubP-GAL80ts, which were then crossed with UAS-TNTa and UAS-TNTi flies. At the permissive temperature (18°C), GAL80ts inhibits GAL4 action,preventing TNT expression, while at 30°C GAL80ts is inactive, allowing GAL4 to drive expression of TNT. To test the effectiveness of this system we first crossed tubP-GAL80ts;p89Da-GAL4 flies to UAS-dsRed flies and raised the offspring at either 18°C or 30°C. The larvae raised at 30°C displayed robust dsRed expression whereas larvae raised at 18°C displayed no dsRed expression in the CNS or PNS, but did retain some expression in the salivary glands, indicating that GAL80ts function was mostly inhibited at this temperature (data not shown). To confirm that we could use this system to regulate TNT expression during adult development we transferred flies containing tubP-GAL80ts;p89Da-GAL4 and UAS-dsRed transgenes from 18°C to 30°C about mid-way through adult development (96 h at 18°C) and examined expression of dsRed in the brain at various times after temperature shift. Examples of these experiments are shown in Fig. 6. No expression was seen in the absence of a temperature shift (Fig. 6A), low levels of expression were seen 7 h after the temperature shift (Fig. 6B) and robust expression was seen 16 h after the shift to 30°C(Fig. 6C).

To determine when TNT needed to be expressed to block eclosion we crossed tubP-GAL80ts;p89Da-GAL4 flies with UAS-TNTa and UAS-TNTi flies and raised the progeny at 18°C until the white puparium (WP) stage. At various times after WP, the animals were transferred to 30°C and the number that subsequently eclosed was noted. The results of this experiment are shown in Fig. 6D. When animals were shifted to 30°C 24–48 h after WP the majority failed to eclose if they expressed TNTa whereas control animals expressing TNTi had a high eclosion rate. By contrast, when animals were shifted later in development, no difference was seen in the rates of eclosion between those expressing TNTa and those expressing TNTi. This suggests that TNTa expression disrupted a developmental event early in adult development that is required for eclosion.

Because TNTa expression in Gyc-89Db neurons blocked the first larval ecdysis it was not possible to determine whether Gyc-89Db neurons were required for all ecdyses. To test this we again used the temperature-sensitive GAL80. Flies were generated that were homozygous for both p89Db-GAL4and tubP-GAL80ts and these were crossed with both UAS-TNTa and UAS-TNTi. The progeny were maintained at 18°C and then shifted to 30°C at various times after WP and the number that successfully eclosed was noted. These data are also shown in Fig. 6D. The results were very similar as those found with Gyc-89Da. When the temperature shift occurred early in adult development, eclosion was subsequently blocked whereas shifting at later stages had little effect on eclosion rates. Interestingly, the critical time in development was also similar to that found for expression of TNTa in the Gyc-89Da neurons. When TNT was expressed earlier than 48 h after WP, the majority of the animals failed to eclose.

We also shifted the p89Db-GAL4;tubP-GAL80ts;UAS-TNTa and UAS-TNTi animals from 18°C to 30°C prior to the WP stage to determine whether they could successfully form puparia. Animals expressing UAS-TNTa that were shifted during the first instar failed to form puparia (2.7% success) whereas those shifted during the second and third instar were more successful (70.5% and 93.8%, respectively). Soon after puparium formation, Drosophilalarvae undergo a process known as head eversion, which appears to be the behavior equivalent to pupal ecdysis in other holometabolous insects(Zdárek and Friedman,1986). All of the larvae containing p89Db-GAL4;tubP-GAL80ts and UAS-TNTa that were shifted to 30°C during the second and third instar that successfully formed puparia also successfully completed head eversion (data not shown). Thus neurons that express Gyc-89Db are required for larval and adult ecdyses, but not head eversion.

In our previous study we used in situ hybridization to show that the atypical sGCs in Drosophila were expressed in a subset of central neurons and peripheral chemosensory neurons(Langlais et al., 2004; Morton et al., 2005a). In the present study we extended these observations by generating fly lines that expressed either GFP or GAL4 under the control of the upstream regions for two of the three atypical sGCs, Gyc-89Da and Gyc-89Db. We first confirmed that the expression patterns generated by the predicted promoter regions for each of the genes matched the expression patterns revealed by in situ hybridization. For the peripheral sensory neurons, we compared the number of cells labeled and their positions in the embryonic in situ preparations with third instar larvae expressing GFP under the direct control of the predicted promoter regions for each gene. In general, the number and position of the cells labeled with either technique matched well. These included neurons innervating the dorsal and terminal organs, lateral and ventral external sensilla and the terminal sensory cones. These data strongly suggested that the sequences of DNA 5′ to the Gyc-89Da and Gyc-89Db genes that we used in our constructs were sufficient to mimic the expression patterns of these genes.

In the earlier in situ hybridization studies, we used the locations of the cells and cell counts to conclude that in most, if not all,of the peripheral sensory neurons all three of the atypical sGCs were co-expressed in the same cells (Langlais et al., 2004; Morton et al.,2005a). Using the GFP and GAL4 reporter lines we now show more clearly the extent of the overlap between Gyc-89Da and Gyc-89Db expression and in contrast to the in situhybridization data, the present study shows relatively little overlap in the expression pattern of these two subunits. There are two primary reasons for the apparent overlap of expression seen using in situ hybridization. Firstly, the sequence similarity between Gyc-89Da and Gyc-89Db makes it difficult to generate gene-specific probes to each gene, possibly resulting in some non-specific labeling. Secondly, most of the cell counts made with the simultaneous application of two probes used a probe to Gyc-88E in combination with either Gyc-89Da or Gyc-89Db. The cell counts were similar in both cases to the cell counts using single probes (Langlais et al., 2004) suggesting that all three genes were co-expressed in the same cells. If we had carried out cell counts combining short,gene-specific probes to both Gyc-89Da and Gyc-89Db, we might have been able to see more cells than with either probe alone, which would have suggested a lack of co-expression. The results from the present study using expression of fluorescent proteins showed that in the olfactory and gustatory neurons there is no overlap in the expression patterns, with each subunit expressed in about four neurons in each ganglion. Along the lateral body wall each of the single neurons innervating external sensilla co-expressed Gyc-89Da and Gyc-89Db whereas the tracheal dendritic neurons only expressed Gyc-89Db. The terminal sensory cones showed a similar pattern to that seen with in situ hybridization– a single neuron in each cone expressed both atypical sGCs. In the CNS it had not previously been possible to examine co-expression patterns because of the number cells and lack of morphological markers. The reporter constructs show clearly that Gyc-89Da and Gyc-89Db are not co-expressed in any CNS neurons. The only locations where they are co-localized are in some of the intrinsic cells of the corpora cardiaca. Because Gyc-89Da and Gyc-89Db are only enzymatically active when co-expressed with the third atypical sGC,Gyc-88E (Langlais et al.,2004; Morton et al.,2005a), we assume that Gyc-88E is also expressed in all the central and peripheral cells that express Gyc-89Da and Gyc-89Db. The in situ hybridization experiments support this, but as we have not been able to define the promoter region for Gyc-88E or generate antisera against Gyc-88E, we have not been able to show this directly.

The identity of the Gyc-89Da and Gyc-89Db cells in the CNS are unknown at this time. We tried a variety of markers that appeared to have similar expression patterns but were unable to see any overlap. These include cells that contain the neuropeptides EH and CCAP and the transcription factor apterous. The transcription factor dimmed is expressed in a wide variety of peptidergic neurons (Hewes et al.,2003) and showed co-expression with a small subset of neurons expressing Gyc-89Da but had no overlap in the CNS with Gyc-89Db. This suggests that this subset of Gyc-89Da neurons is peptidergic, but the identity of the peptide is unknown.

We used the neurotoxin TNT to determine the function of the neurons that express Gyc-89Da and Gyc-89Db. TNT blocks synaptic transmission by cleaving the synaptic vesicle protein synaptobrevin(Schiavo et al., 1992; Link et al., 1992) and expression of TNT in specific neurons in Drosophila has helped define neuronal networks involved in a variety of behaviors(Martin et al., 2002). When we expressed TNT in either the Gyc-89Da or Gyc-89Db neurons we found that the animals died at three specific stages (Fig. 3). TNT expression in Gyc-89Da neurons caused about half of the animals to die at the late larval stage and the remainder of the animals to die just prior to the onset of the eclosion motor program. The lethality of third instar larvae was associated with burrowing into the food and could be rescued by allowing the animals to develop in a thin layer of food. We have shown that the activity of the atypical sGCs is regulated by oxygen and have suggested that they could act as oxygen sensors(Morton, 2004b). If the Gyc-89Da neurons are excited by reduced oxygen levels and activate hypoxia-associated behaviors, when synaptic transmission of these neurons is blocked by TNT the animals would no longer respond to hypoxia. One response of larvae to hypoxia is to withdraw from the food and activate searching behaviors (Wingrove and O'Farrell,1999). Our results are consistent with the Gyc-89Da neurons either detecting hypoxia or activating the hypoxia escape behavior. If synaptic transmission is blocked in these neurons the animals would fail to respond to hypoxia, continue to burrow into the food and suffocate. Most of the larval lethality that we observed was late in the third instar, presumably because at this time locomotion increases as the larvae crawl out of their food to seek a location to pupate. Wingrove and O'Farrell(Wingrove and O'Farrell, 1999)have shown that the hypoxia escape response is probably mediated by a cGMP/protein kinase G (PKG) pathway and we have preliminary data that show that cGMP in the Gyc-89Da neurons is necessary for a normal hypoxia escape response (Morton et al.,2005b).

The other major period of lethality for animals when TNT was expressed in the Gyc-89Da neurons was just prior to adult eclosion. All of the stages that precede the onset of the eclosion motor program occurred normally and in the appropriate order until the final stage – head expansion. This time corresponds closely to the time at which defects in eclosion were first detected in flies lacking the neurons that contain EH(McNabb et al., 1997), a neuropeptide known to trigger ecdysis and eclosion in most, if not all,insects (Truman et al., 1981). The simplest interpretation of our findings is that Gyc-89Da is expressed in EH neurons and expression of TNT in these cells prevented EH release. This is not the case as there was no overlap in EH and Gyc-89Da expression patterns. Cells that are immediately upstream and downstream of EH in the pathway that regulates eclosion are the Inka cells of the epitracheal glands and the CCAP neurons (see Ewer and Reynolds, 2002; Clark et al., 2004). No overlap in the expression pattern of Gyc-89Da with either of these cells was seen. It is possible that the Gyc-89Da neurons act further upstream of the EH cells, although the Gyc-89Da neurons must affect a larger number of neurons associated with eclosion than the EH cells as the phenotype of TNT expression in the Gyc-89Da neurons is much more severe than the EH cell deletion. Almost all the p89Da-TNTa animals failed to eclose whereas about a third of the flies lacking EH cells complete development and eclose(McNabb et al., 1997). As with the EH cell deletions, expression of TNT in the Gyc-89Da neurons prevented normal tanning and wing expansion.

Expression of TNT in the Gyc-89Db neurons also phenocopied the effects of eliminating a neuropeptide associated with eclosion. ETH acts both upstream and downstream of EH as part of a positive feedback loop(Clark et al., 2004) and mutations that eliminate the ETH gene cause lethality at the first larval ecdysis (Park et al.,2002). The timing of death in ETH–/– larvae(Park et al., 2002) is very similar to that of larvae expressing TNT in the Gyc-89Db neurons – just prior to air filling of the tracheae. In wild-type larvae it only takes about 2 min between tracheal collapse and air filling of the tracheae(Park et al., 2002; Clark et al., 2004). Interestingly, in the EH cell deletions, many of the larvae also failed to air fill their tracheae and this was a major cause of lethality(McNabb et al., 1997). Thus elimination of either ETH or EH results in a similar phenotype to blocking neurotransmission in the Gyc-89Db neurons. As described above for the Gyc-89Da neurons, a simple explanation for this is that TNT is expressed in ETH, EH or CCAP cells preventing the release of the peptides. This is not the case as we saw no overlap in expression patterns with any of these peptides and Gyc-89Db. Additional peptides believed to be involved with activation of the ecdysis behavior program include peptides in the apterous cells(Park et al., 2004), corazonin(Kim et al., 2004) and the products of the hugin gene (Meng et al., 2002). Although we saw no overlap with apterous it is possible that the release of other peptides is prevented by expression of TNT in the Gyc-89Da or Gyc-89Db neurons. Although TNT blocks synaptic transmission at conventional synapses in Drosophila, there is little experimental support that in Drosophila TNT will block peptide release(Martin et al., 2002).

Another possibility is that synaptic transmission in Gyc-89Db neurons is directly required for successful air filling of tracheae. The physiological basis for air filling the tracheae at each molt is not known although the ETH neuropeptide is involved (Park et al.,2002). Gyc-89Db is expressed (in the absence of Gyc-89Da) in sensory neurons that are closely associated with tracheae (Fig. 2B). If these neurons signal to the CNS when apolysis has progressed sufficiently to trigger air filling, failure of synaptic transmission in these cells could account for the inability of the animal to successfully fill its tracheae with air.

The data discussed above support the model that synaptic transmission in neurons that express Gyc-89Da and Gyc-89Db is required for either the activation or the generation of the eclosion and ecdysis motor programs. This model predicted that TNTa expression in these neurons just prior to ecdysis or eclosion is sufficient to block the behavior. Our experiments using GAL80ts to control the temporal expression of TNTa showed that this was not the case. TNTa expression about 100 h before eclosion is required to block eclosion. This delay is not explained by the time it takes for the inhibition of GAL4 activation to be removed. McGuire et al.(McGuire et al., 2003) found that 30 min after the temperature shift a reporter gene was detectable and half-maximal expression was seen after 3 h. Our data showed that expression of the red fluorescent protein dsRed was detectable 7 h after the temperature shift and was robust 16 h later.

Our findings suggest that TNT expression probably blocks a developmental event that is required for the subsequent expression of eclosion behaviors. Previous studies have also shown developmental effects of TNT expression in Drosophila neurons. In the embryonic neuromuscular system, expression of TNT in motor neurons resulted in the expected blockage of neuromuscular synaptic transmission and also reduced synaptic input to the motor neurons(Baines et al., 1999). In a detailed analysis of the effects of TNT expression in the developing optic lobe, Hiesinger et al. (Hiesinger et al.,1999) found that cell survival, axonal pathfinding and target recognition were unaffected by TNT expression, but morphological alterations in photoreceptor terminals was observed between 25% and 50% of adult development. This time corresponds to the window of sensitivity that we observed for the blockage of adult eclosion. This suggests that expression of TNT in the Gyc-89Da and Gyc-89Db neurons disrupts their connectivity and leads to a blockade in eclosion. It is interesting that the Gyc-89Da neurons are not required for larval ecdysis, whereas the Gyc-89Db neurons are, but for adult eclosion TNT expression in either population of cells blocked the behavior. In the larval CNS there was very little overlap of expression, and similarly in the adult CNS few, if any, neurons co-expressed Gyc-89Da and Gyc-89Db.

In summary, by using targeted expression of TNT in neurons that express two atypical soluble GCs, we have shown that synaptic transmission in neurons that express Gyc-89Da is required for adult eclosion and neurons that express Gyc-89Db are required for larval ecdysis and adult eclosion. By contrast, neither of these populations of neurons is required for head eversion. The use of the driver lines also enabled us to more completely map the expression patterns of these two GC subunits and we found some co-expression in peripheral sensory neurons, but almost no overlap in CNS expression.