Retrograde Gbb signaling through the Bmp type 2 receptor Wishful Thinking regulates systemic FMRFa expression in Drosophila

Neuropeptides are a highly diverse group of molecules that modulate specific behaviors and physiological responses in a wide variety of phyla ranging from cnidarians to humans. Although several families are recognized on the basis of conserved sequences, among the most broadly distributed are FMRFamide-related peptides (FaRPs) (Geary et al., 1999; Grimmelikhuijzen et al., 1989; Hinuma et al.,2000; Yang et al.,1985). These peptides share a common C-terminal RFamide sequence and in Drosophila are produced from at least five genes, dromyosuppressin, drosulfakinin, two neuropeptide F-like genes and FMRFa (reviewed by Benveniste and Taghert, 1999; Merte and Nichols, 2002). Some of these loci can produce several different peptides exhibiting unique N-terminal sequences by selective proteolytic processing of a proprotein. Sequence analysis of the FMRFa protein suggests that it encodes at least eight unique peptides(Nambu et al., 1988; Schneider and Taghert, 1988; Schneider and Taghert, 1990),three of which have been recovered from tissue extracts(Nambu et al., 1988; Nichols, 1992; Schneider and Taghert, 1988; Schneider and Taghert, 1990). In most insects, FMRF-type peptides mediate their neuromodulatory effects by stimulation of G protein-coupled receptors(Cazzamali and Grimmelikhuijzen,2002; Chiba et al.,1992; van Tol-Steye et al.,1999; Volterra and Siegelbaum,1988), but in snails(Cottrell, 1997) and mammals(Lingueglia et al., 1995)there is evidence that they can also directly activate ligand-gated channels.

The physiological processes that FaRPs influence are quite varied. The original invertebrate peptide was isolated from clam ganglia and exhibited marked cardioexcitatory activity (Price and Greenberg, 1977). Other family members, such as Myosuppressins and Sulfakinins, affect spontaneous contractions of the visceral and oviduct muscles and thereby regulate feeding and egg-laying behaviors(Lange and Orchard, 1998; Nachman et al., 1986; Wang et al., 1994; Wang et al., 1995a; Wang et al., 1995b). In Aplysia, these peptides can influence learning and memory(Guan et al., 2002; Mackey et al., 1987; Small et al., 1989) while in vertebrates endogenous FaRPs regulate analgesic effects of opiate peptides and influence the electrical activity of some central brain synapses(Askwith et al., 2000; Gayton, 1982; Kavaliers, 1990; Kavaliers and Yang, 1991; Nishimura et al., 2000; Tang et al., 1984; Yang et al., 1985).

In invertebrates, FaRPs can also influence body wall muscle activity. Early studies in the locust suggested that FMRF-like peptides enhance synaptic transmission at the neuromuscular junction(Robb and Evans, 1994), and more recent studies in crustaceans suggest that these peptides modulate presynaptic Ca²⁺-channel activity(Rathmayer et al., 2002). In Drosophila, peptides produced from the FMRFa gene can also enhance synaptic efficacy at the neuromuscular junction when perfused onto standard larval nerve-muscle preparations(Hewes et al., 1998). Such treatment results in a significant increase in muscle contraction or twitch tension. Drosophila FMRFa is expressed in many neurosecretory cells,including the Tv neurons that innervate a specialized tissue known as the neurohemal organ (NHO). As the NHO releases products into the hemolymph, FaRPs probably act systemically in a hormone-like fashion to regulate NMJ synaptic activity in vivo (O'Brien et al.,1991; Schneider et al.,1991; Schneider et al.,1993a; Schneider et al.,1993b). Elucidating the mechanisms that control the peptidergic phenotype and activity of particular neurosecretory cells is therefore important for understanding how NMJ activity may be modulated by the neuroendocrine system.

Within the vertebrate nervous system, members of the bone morphogenetic protein (Bmp) subgroup of TGFβ and their cognate receptors have been implicated in controlling several different aspects of neural development and function, including neurulation, morphogenesis, lineage decisions and cellular maturation (reviewed by Mehler et al.,1997). Bmps have also been implicated in neural specification processes. For example, the induction and maintenance of the neuronal cholinergic phenotype in the central nervous system is influenced by Bmp9(Lopez-Coviella et al., 2000),and trunk neural crest cells are induced to an adrenergic phenotype by Bmp2,Bmp4 and Bmp7 (Reissmann et al.,1996; Varley and Maxwell,1996). Other Bmps have also been implicated in regulating neurotransmitter expression in sympathetic(Fann and Patterson, 1994; Lo et al., 1998; Schneider et al., 1999),spinal cord (Kalyani et al.,1998), mesencephalic (Jordan et al., 1997; Reiriz et al.,1999), striatal (Hattori et al., 1999) and serotonergic neurons(Galter et al., 1999). In most of these cases, the Bmp is not required for the differentiation of these neurons, rather it helps them obtain their final phenotypic characteristics by inducing the expression of genes specific for the function of that neuron. In C. elegans, daf-7, a TGFβ-type ligand, is required for maintaining, but not initiating, the expression of chemoreceptors in sensory neurons, thus modulating the chemosensory properties of specific neurons(Nolan et al., 2002).

With these examples as precedent, we sought to determine whether Bmp signals might influence the expression of neuroendocrine phenotypes in Drosophila. We and others have recently described a novel Drosophila Bmp type 2 receptor, coded for by the wishful thinking (wit) locus, that is primarily expressed in, and required for, proper nervous system function(Aberle et al., 2002; Marqués et al., 2002). Mutations in wit result in pharate lethality caused, in part, by defects in the growth and physiology of motoneuron synapses. We show here that mutations in wit also affect the peptidergic phenotype of certain FMRFa-expressing cells. In particular, we find that FMRFaexpression is eliminated in the Tv neurons that contribute to the systemic supply of FMRFa peptides through release at the neurohemal organ. We show that the regulation of FMRFa expression in Tv neurons is mediated by the Bmp ligand Gbb, as gbb null mutations also eliminate FMRFaexpression in Tv neurons. Furthermore, we demonstrate that supplying Gbb to the dorsal neurohemal cells restores FMRFa expression in Tv neurons. As Tv neuron axons arborize onto the neurohemal cells, this strongly suggests that Gbb signals in a retrograde manner to specify the peptidergic phenotype of Tv neurons. Consistent with this view, we find that overexpression in neuroendocrine cells of Dynamitin or a dominant-negative form of p150/Glued,both components of the Dynactin/Dynein motor complex, also eliminates FMRFa expression in the Tv neurons. Finally, we show that providing FMRFain neuroendocrine cells using the Gal4/UAS system partially rescues the lethal phenotype of wit mutants, even though they still exhibit structural and physiological synaptic defects. We suggest that Bmp signaling provides a global cue that not only regulates the growth of the NMJ synapses locally(Aberle et al., 2002; Marqués et al., 2002; McCabe et al., 2003) but also controls their systemic modulation by the neuroendocrine system.

The wit^A12, wit^B11 and the small deficiency Df(3L)C175 have been described previously(Marqués et al., 2002). Unless otherwise stated, in all mutant characterizations and rescue experiments wit mutants refers to the null heteroallelic combination wit^A12/wit^B11. The gbb¹, gbb² and gbb⁴ alleles are described elsewhere (Wharton et al.,1999) as are the sax alleles(Twombly et al., 1996). The elav>Gal4 and G14>Gal4, lines were obtained from C. Goodman. The C929 and 386 Gal4 lines were obtained from P. Taghert and 24B>Gal4 was obtained from the Bloomington stock center. The UAS-Dynamitin stocks are described elsewhere (Duncan and Warrior,2002) as is UAS-ΔGl(Reddy et al., 1997). All germline transformations were carried out by injection into Df (1) w,y^87c27 stock using standard protocols.

UAS>FMRFa was generated by synthesis of two 100 bp complementary oligonucleotides that encode the last few amino acids of the FMRFagene. After formation of duplexes, this oligo was inserted into the PstI-SpeI sites of an incomplete FMRFa cDNA(Schneider et al., 1991) to generate a full-length coding region. The full-length FMRFa gene was excised from the pBluescript clone as an EcoRI-SpeI fragment and inserted into the EcoRI-XbaI sites of pUAST(Brand and Perrimon, 1993).

pCasper-Ubiquitin (pUC) (Brummel et al.,1994) was digested with EcoRI, blunted with Klenow, and religated. The plasmid was cut with XbaI and StuI, and a new polylinker was added. Unique sites in the polylinker: 5′ NotI, SacII, HpaI, AvrII, EcoRI, XbaI 3′.

A StuI site (AGGCCT) was inserted into the middle of the sax transmembrane domain in pBluescript by site-directed mutagenesis. A NotI fragment was inserted into pUC.

• Wild-type sequence: GGT CCT TTT CTG GTC ATC

• Mutant sequence: GGT CCT TTA GGC CTC ATC

• Wild-type protein: GPFLVS

• Mutant protein: GP LGLS

• Bold letters indicate changed residues.

A StuI site (AGGCCT) was inserted into the middle of the transmembrane domain of the tkv1 isoform in pBluescript by site-directed mutagenesis. A NotI fragment was inserted into pUC.

• Wild-type sequence: ATC ATC ATC TCC CTG TCC

• Mutant sequence: ATC ATC ATA GGC CTG TCC

• Wild-type protein: IIISLS

• Mutant protein:IIIGLS

StuI (AGGCCT) and HpaI (GTTAAC) sites were inserted into the middle of the wit transmembrane domain in pBluescript by site-directed mutagenesis. A HindIII (blunted)-HpaI fragment was inserted into the HpaI site of pUC2.

A StuI-XbaI fragment of pBluecript-wit5′(StuI, HpaI) was replaced from pBluecript-wit5′(StuI, HpaI)

• Wild-type sequence: CTT GCC GGT GGA CTC ACA GCC CTC ACA ATC GGC

• Mutant sequence: CTT GCA GGC CTA CTC ACA GCC CTC ACA ATC GGC

• Wild-type protein: LAGGLTALTIG

• Mutant protein: LAGLLTALTIG

• Wild-type sequence:ATC TTC CTG GCT GTT CAA TAT

• Mutant sequence:ATC TTC CTG GCT GTT AAC TAT

• Wild-type protein: IFLAVQY

• Mutant protein: IFLAVNY

The chimeric proteins are fused in the middle of the transmembrane domains.

• pUC2-sax/wit: a NotI-StuI fragment of pUC2-wit was replaced by the extracellular domain fragment of pUC-sax (StuI).

• pUC2-tkv1/wit: a NotI-StuI fragment of pUC2-wit was replaced by the extracellular domain fragment of pUC-tkv1 (StuI).

• pUC2-wit/sax: a StuI-XbaI fragment of pUC2-wit5′was replaced by the sax transmembrane and kinase domain fragment.

• pUC2-wit/tkv: a StuI-XbaI fragment of pUC2-wit5′was replaced by the tkv transmembrane and kinase domain fragment.

The following antibodies were used at the indicated dilutions for characterization of the wit and gbb mutant phenotypes:monoclonal anti-lacZ (Promega) 1/1000; rabbit anti-PSMAD1 (ten Dijke), 1/1000 in embryos and 1/500 in larvae; monoclonal anti-Csp (Zinsmaier)1/400; and rabbit anti-FMRFamide (Peninsula Laboratories) 1/1000. The Alexa series (Molecular Probes) of secondary antibodies were used for immunofluorescence at 1/500 dilution. Larvae were dissected and fixed in 4%formaldehyde in PBS containing 0.5 mM EGTA, 5 mM MgCl₂ for 10-20 minutes at room temperature for pMad and lacZ staining. For FMRFamide staining, a protocol was used in which fixation takes place in 0.1 M NaPO₄, 0.3% Triton X-100, 0.1%sodium azide, 0.1% BSA for 2-6 hours(McCormick et al., 1999). Staining of all tissues was visualized in a Zeiss Axioplan2 with a CARV unit for confocal microscopy.

The standard third instar larval body-wall muscle preparation was used for electrophysiology as previously described(Jan and Jan, 1976; Zhang et al., 1998). Briefly,excitatory junctional potentials (EJPs) in muscles 6 and 7 were elicited by stimulating the innervating motor nerve bundle with a suction electrode in HL-3 solution containing 1 mM CaCl₂. The recording microelectrode had an input resistance between 15 MΩ and 25 MΩ. EJPs were acquired and digitized using a PC computer with the use of pCLAMP 8 software(Axon Instruments). The analysis and presentation of figures were conducted on Clampfit (Axon Instruments) and Origin (Origin lab). Samples used for final analysis were obtained from at least five different larvae. ANOVA and Unpaired Student's t-test were used for data treatment (mean±s.e.). The resting potential of the muscles was between –65.8±1.2 mV for w control, –64.9±0.6mV for wit^A12/D(3)C175, and–63.9±1.6 mV for the FRMFa rescue. They were not statistically different from each other (P>0.5).

We previously reported that wit mutant flies die during the pharate adult stage and never escape the pupal case. However, some of these individuals exhibit normal peristaltic movements typical of pre-eclosion and,when dissected from the pupal case, are capable of limited spastic movements and will live for several days. During this time we observed that they never inflate their wings, a phenotype that has been ascribed to neuroendocrine defects (Clark et al., 2002; McNabb et al., 1997). As FMRFamide neuropeptides have been shown to affect both synaptic efficacy and muscle twitch tension, and as Bmps have been implicated in vertebrate systems as regulators of final neural phenotypes, we examined whether loss of wit signaling disrupted FMRFa expression. Initially, we employed an antibody derived against the peptide FMRFamide. This antibody likely recognizes several classes of FaRPs. As shown in Fig. 1 (compare A with B), we find that in crawling third instar wit mutants there is a specific disruption in the accumulation of FaRPs in the Tv neurons and the neurohemal organ. Other FaRP-positive cells such as the subesophageal neurons (SE2)(Benveniste and Taghert, 1999)in the ventral ganglia and several neurons in the brain appear to express FaRPs normally or are only moderately affected. It is important to note that the NHO is composed of two components: the two or three NHO cells and the large varicosities formed by the Tv neurons axons that terminate on the NHO cells. FMRFa accumulation in the Tv neurons appears to be primarily in the large varicosites because, at the RNA level, FMRFa does not appear to be expressed in the NHO cells themselves(O'Brien et al., 1991).

The loss of FaRP staining in the Tv neurons and the NHO varicosities does not necessarily mean that Wit signaling affects transcriptional activation of a FaRP in the these neurons since the effect could potentially occur at several other levels, including production of necessary processing enzymes(Jiang et al., 2000; Kolhekar et al., 1997; Renn et al., 1998). As the FMRFa gene is known to be strongly expressed in the Tv neurons that innervate the NHO (O'Brien et al.,1991; Schneider et al.,1993a; Schneider et al.,1993b), we examined the effect of wit mutations on the expression of a FMRFa/lacZ transgene. The regulatory region of FMRFa comprises several different enhancers, including one that produces strong expression in the Tv neurons(Benveniste and Taghert, 1999). As shown in Fig. 1C-F, loss of wit signaling specifically eliminates expression of the FMRFa/lacZ transgene in the Tv neurons and the NHO. Expression in the subesophageal neurons (Fig. 1D), as well as several other neurons in the brain, remains unaffected (data not shown).

As wit expression is observed in many neuronal cells, we wished to determine if it is required directly in Tv neurons or whether it might function indirectly via an interneuron signal. To address this issue, we employed three different Gal4 lines that express Gal4 in either all or specific subsets of neuronal cells. The elav driver is expressed in most differentiated neurons and, supplying wit with this driver restores FaRP expression in the Tv neurons of wit mutants(Fig. 2A). The C929 and 386 lines express Gal4 in much more limited sets of neuroendocrine cells(Taghert et al., 2001),including the Tv neurons but not the NHO (for C929 see Fig. 2C-F). As these drivers also rescue FaRP accumulation in the Tv neurons and the NHO varicosities (2B),we conclude that wit is required only in the Tv neurons and is not required to mediate an interneuron signal, nor is it required in NHO cells. In order to confirm that Wit is required in Tv neurons, we took advantage of the variable expression of the OK6 driver in Tv neurons. We have noticed that in addition to the previously described motoneuron expression (Aberle, 2002),this driver is also expressed in a random fashion in some or all of the Tv neurons. We looked at FMRFa expression in wit mutants rescued with a witGFP transgene driven by OK6>Gal4. We find that FMRFa expression is variably recovered in different animals, but in all cases (five animals, 17 Tv neurons) the expression of FMRFacorrelates with the expression of WitGFP in those neurons(Fig. 3). In these experiments,we could also detect WitGFP in the NHO (not shown). In no case did we see WitGFP expression in Tv neurons without FMRFa expression.

The absence of systemic FMRFamide could be due to the absence of the Tv neurons or to the lack of proper transcriptional activation of the FMRFa gene. Expression of FMRFa in Tv neurons is under the control of the transcription factor Apterous (Ap), that directly binds to upstream regulatory elements (Benveniste et al., 1998). We reasoned that Ap activity might be regulated by Drosophila Smads directly downstream of Wit (see below). To test this hypothesis we examined ap expression in wit mutants. Fig. 4 shows the expression of a GFP reporter driven by ap>Gal4 in a wild type(Fig. 4A-F) or a witmutant background (Fig. 4G-L). In addition to many other cells, ap is expressed in three clusters of four neurons on each side of the ventral ganglion(Fig. 4A,D)(Benveniste et al., 1998). One of these neurons is the FMRFa-expressing Tv neuron that innervates the NHO (Fig. 4C,F). In wit mutants, FMRFamide expression is absent in the Tv neurons (I),but ap expression remains normal(Fig. 4G,J-L). This suggests that ap is not a transcriptional target of Wit and also shows that Tv neurons are present in the ventral ganglia of wit mutant larvae. This conclusion is further supported by the fact that the NHO proper is still present in wit mutants and is roughly the same size as in wild type(Fig. 1E-H). Because much of this organ comprises large varicosities from the Tv axons that arborize onto the NHO cells, it appears that these arbors do form. As with the C929 driver,we could also rescue FMRFa expression in wit mutant Tv neurons using ap>Gal4 and UAS-wit (not shown).

We have previously shown that Wit acts to transduce Bmp-type signals(Aberle et al., 2002; Marqués et al., 2002). This requires the Bmp type I receptors Tkv and Sax and results in accumulation of phosphorylated Mad (P-Mad) in motoneurons(McCabe et al., 2003; Rawson et al., 2003). However,recently Wit has been found to be redundant with Punt, a second Drosophila Bmp/Activin type II receptor, for transducing activin-like signals through Smad2 in mushroom body neurons(Zheng et al., 2003). In this case, Wit appears to signal in conjunction with Baboon, an activin-like type I receptor (Brummel et al.,1999). To distinguish whether Wit transduces a Bmp or an Activin type signal to Tv neurons, we initially examined whether P-Mad accumulates in Tv neurons. As shown in Fig. 5A,B), we found clear evidence that these neurons do indeed accumulate P-Mad, suggesting that they receive a Bmp type signal. In a wit mutant background, P-Mad accumulation in all ventral ganglion neurons including the Tv neurons is eliminated(Fig. 5C,D).

We also wished to test which type I receptor is required for this process. We found that in the strong heteroallelic sax³/sax^Df(2)Hr-1 combination, mutant larvae show greatly reduced expression of FaRPs in the Tv neurons and the NHO(Fig. 5E). As animals containing strong tkv alleles do not survive beyond the embryonic stage, we could not test directly for the requirement of Tkv in this process. As expected from the model for signal transduction by TGFβ receptors, a functional receptor complex can be reconstituted by simultaneous expression of two reciprocal chimeras, one containing the extracellular ligand binding domain of the type I receptor and the intracellular kinase domain of the type II receptor and the other formed by the extracellular domain of the type II receptor and the intracellular domain of the type I receptor(Feng et al., 1995). This functional complementation is a strong indication that the type I and type II receptors that make up the chimeras act as partners in vivo. We found that the requirement for Wit can be circumvented when a chimeric pair of receptors, one containing the extracellular domain of Tkv fused to the intracellular domain of Wit, and the other composed of the extracellular domain of Wit fused to the intracellular domain of Tkv (see schematic in Fig. 5), are co-expressed in a wit mutant background. By themselves, these receptors are unable to rescue either NMJ defects or FMRFa expression (data not shown). However when both are expressed using ubiquitin promoters, then we observe rescue of FaRP expression (Fig. 5F), strongly supporting that Tkv acts in vivo with Wit to activate FMRFa expression in Tv neurons. An equivalent set of chimeras of Sax and Wit rescues neither defect (data not shown). This observation is consistent with our previous finding that ubiquitous expression of Tkv alone can rescue sax mutants, while ubiquitous Sax cannot rescue tkv mutants (Brummel et al., 1994). Thus, we conclude that, although Sax is required for both NMJ growth and FMRFa expression, it is not able by itself to transduce sufficient signal for these two processes. Instead these data strongly support the notion that Wit is supplying a Bmp signal through both Tkv and Sax.

Although these findings demonstrate that reception of a Bmp signal is an essential component of FMRFa gene regulation in Tv neurons, they do not rule out that an activin-type signal may also be received and required by the Tv neurons to stimulate FMRFa expression. To probe this issue, we examined FMRFa expression in baboon mutant third instar larvae. As shown in Fig. 5G, we found that expression of the FMRFa/lacZ transgene is still evident in babo mutant larvae, demonstrating that an activin-type signal is not required for FMRFa expression.

The Drosophila genome contains seven TGFβ type ligands. Three of these, Dpp, Screw and Gbb, have been shown to transduce Bmp-type signals(Mad) and to use the type I receptors Tkv and Sax(Brummel et al., 1994; Haerry et al., 1998; Nguyen et al., 1998). Two others, Activin and Activin-like protein, transduce signals through Smad2(T.E.H. and M.B.O., unpublished). The signaling pathways used by Maverick and Myoglianin remain untested. Among the three Bmp-type ligands, Gbb seemed a likely candidate for controlling expression of FMRFa as it is broadly expressed, at least in embryos (Doctor et al., 1992; Wharton et al.,1991), and can signal through Wit to regulate P-Mad accumulation in motoneurons and tissue culture cells(McCabe et al., 2003). As shown in Fig. 6A, gbbis strongly expressed in the larval brain lobes and much more weakly in the ventral ganglia. Interestingly, we also note that gbb shows enriched expression in the NHO relative to other ventral ganglia neurons(Fig. 6A). Thus, Gbb is expressed in the correct place to be a FMRFa regulating ligand.

To test directly whether gbb is required for FMRFaexpression, we examined gbb mutant larvae for FaRP expression. As shown in Fig. 6C,D, we found that, although the gbb¹/gbb⁴ hypomorphic combination still expresses FaRPs, the stronger gbb¹/gbb² mutant larvae show a selective loss of FaRP expression in Tv neurons and the NHO, strikingly similar to that exhibited by wit mutants. When Gbb is re-supplied to all neurons using the pan neuronal elav>Gal4 driver(Fig. 6E), expression of FaRP is restored; if the muscle specific G14>Gal4 driver is utilized, only weak expression is occasionally seen in some Tv neurons(Fig. 6F).

The fact that gbb is expressed in the NHO onto which the Tv neuron axons terminate suggested that perhaps Gbb signals to these neurons via a retrograde mechanism. To address this issue, we took advantage of the observation that the 24B>Gal4 driver, while expressed in muscles, is also specifically expressed in the NHO cells(Fig. 7D-F) but not in the Tv neurons themselves (Fig. 7A-C). When 24B>Gal4 is used to express Gbb in the NHO, we found that FaRP expression is restored in the Tv neurons(Fig. 7G), while as shown above, use of the muscle specific driver G14>Gal4 does not rescue(Fig. 6F).

Although these results strongly suggest that Gbb can signal in a retrograde fashion from the NHO to control expression of FMRFa in the Tv neurons, they do not directly examine whether this is the case in vivo. To address this issue, we overexpressed a truncated form of p150/Glued(ΔGl) in neuroendocrine cells. Overexpression of this protein interferes with assembly of the Dynein retrograde motor with its cargo leading to the development of `roadblocks' in which transported proteins collect at jams within axons (Martin et al.,1999). These jams can ultimately interfere with both anterograde and retrograde transport. When we overexpress ΔGl in neuroendocrine neurons using the C929>Gal4 driver, these animals survive to the pupal stage and about 50% eclose. However, many eclosed adults exhibit uncoordinated movements and some do not inflate their wings(Fig. 9E) reminiscent of wit mutants (not shown). When we examine these larvae for FaRP expression, we find that it is dramatically reduced in the Tv neurons as well as the NHO relative to wild type (Fig. 8A). When the 24B>Gal4 line is used to express ΔGl in the NHO, but not in Tv neurons, then no effect on FaRP expression is seen(Fig. 8B). In these experiments, other FaRP-positive cells in the brain show equivalent staining in both wild type and the ΔGl-expressing line.

These results indicate that the Tv neurons are particularly sensitive to a block in Dynein function consistent with a requirement for retrograde transport in specifying their peptidergic phenotype. To ensure that this is not an artifact of blocking another function of P150/Glued, we also overexpressed p50/Dynamitin. In mammals, overexpression of the Dynamitin subunit of the Dynactin complex has been shown to dissociate Gl from the Dynactin complex, thereby disrupting cargo attachment to the Dynein motor(Echeverri et al., 1996; Eckley et al., 1999). In Drosophila, phenotypes consistent with Dynein motor blocks have also been seen when Dynamitin is overexpressed(Duncan and Warrior, 2002; Januschke et al., 2002). When we use C929>Gal4 to overexpress Drosophila Dynamitin in Tv neurons, we see a similar loss of FaFR expression as was found for ΔGl(Fig. 8C).

To demonstrate that these effects are not the result of interference with P-Mad transport from the cell cytoplasm to the nucleus, we also expressedΔGl in the embryonic gut using the Y45>Gal4 driver(Wharton et al., 1999). We have previously shown that P-Mad accumulates extensively in the midgut beginning at stage 15 of embryogenesis(Marqués et al., 2002). Overexpression of ΔGl did not interfere with P-Mad accumulation in the midgut (Fig. 8D versus 8E), nor did it interfere with P-Mad accumulation in dorsal cells when a ubiquitous Gal4 driver such as daughterless is used(Fig. 8F versus 8G). The only tissue showing a loss of P-Mad staining when using the ubiquitous da>Gal4 line is the CNS(McCabe et al., 2003). We conclude that P-Mad accumulation in neurons is most sensitive to disruption of Dynein motor function, consistent with the notion that Bmp signaling in these cells requires retrograde transport of at least one component as part of its signal transduction mechanism.

As FaRPs are known to enhance synaptic transmission at the neuromuscular junction, we wished to determine if re-supplying FMRFa alone could partially ameliorate the wit mutant phenotype. We have previously shown that wit mutants exhibit structural defects at active zones and are defective in synaptic transmission(Aberle et al., 2002; Marqués et al., 2002). Our rationale was that, although re-supply of FaRPs would probably not rescue the primary structural defect of the synapses, it might enhance synaptic transmission enough to allow some animals to eclose. For this purpose we constructed a UAS>FMRFa transgene and expressed it in neuroendocrine cells using C929>Gal4. At 25°C we find that ∼80%(n=65) of wit mutant pharate adults can partially escape from the pupal case (Fig. 9B). In the absence of FMRFa expression less than 0.5% of witmutants (0 of 212) open the operculum. At 28°C about 30% of the females can fully eclose (Fig. 9C) and some inflate their wings (Fig. 9D). While these animals move fairly well they usually die within several days without producing progeny. Although many of these animals are able to eclose, we see no rescue in the size of the synapse(Fig. 9F,G) nor is the primary defect in synaptic transmission rescued(Fig. 9H,I). We conclude therefore, that the inability of wit mutant animals to eclose is the result of at least two defects, one in NMJ synaptic growth and neurotransmitter release, and the other in the production of systemic FaRPs that act as neuromodulators at the NMJ.

Maintaining constant synaptic efficacy at the neuromuscular junction throughout development requires a precise coordination of muscle and synapse growth. In addition, NMJ function is likely to be modulated by both behavioral and physiological inputs that derive from stage specific developmental signals and changes in the environment. We have previously shown that Bmp signaling through the type II receptor Wit is a key regulator of NMJ synaptic growth(Aberle et al., 2002; Marqués et al., 2002). More recently, we have found that the Bmp-like ligand Gbb signals through Wit to mediate synapse growth (McCabe et al.,2003). In this report, we show that the Wit pathway specifically turns on the systemic component of FMRFamide in Tv neurons. Absence of either the ligand Gbb or the type II receptor Wit results in loss of expression of the FMRFa precursor protein in Tv neurons.

In Drosophila FMRFamide peptides have been shown to enhance synaptic transmission and muscle twitch tension when perfused onto standard larval nerve-muscle preparations (Hewes et al., 1998); however, their in vivo role(s) are not known as no mutations in the FMRFa gene have been identified. As with most neuropeptides, FaRPs are thought to act as neuromodulators and neurohormones. The Tv-produced FaRPs are released into the hemolymph through the neurohemal organ and hence are able to act on every tissue in the animal that is not blocked to hemalymph contact. We have previously hypothesized that the lethality of wit mutants is due to the lack of proper synaptic transmission at the NMJ, resulting in the animals not being able to eclose from the pupal case (Marqués et al., 2002). The lack of systemic FMRFamide described here would be expected to further decrease synaptic efficiency and the ability of wit mutants to eclose. The fact that loss of FMRFa does contribute to the lethal phenotype is supported by the partial rescue of witmutants by overexpression of FMRFa. These results are consistent with the view that in vivo, FMRFa peptides probably enhance NMJ synaptic activity similar to their in vitro documented effects on standard larval electrophysiological preparations.

It is important to note that although the lethal phenotype is partially reversed, the morphological and physiological synaptic defects reported for wit mutants are not rescued by overexpression of FMRFa. The simplest interpretation is that the excess of FaRPs enhances the efficiency of wit mutant synapses in vivo without correcting the underlying developmental defects. Although one might expect a significant improvement of the electrophysiological phenotype, this is not detected(Fig. 9), probably because the excess FaRPs are either washed off the preparation during standard dissection prior to recording or act for only short periods.

Although we have argued that the FMRF rescue effect is the result of enhance synaptic transmission at the NMJ, we can not exclude the possibility that the rescue of eclosion by FMRFa expression in witmutants is due to a central effect on ecdysis regulation, and not to a peripheral effect on the NMJ. The driver used, C929, expresses FMRFain all neurosecretory cells, so extra FMRFamide is conceivably produced in the CNS. However, it is important to recognize that wit mutants are defective only in the systemic, Tv-secreted form of FMRFamide. Although FMRFamide has a well-studied effect in modulating visceral and somatic muscle contraction (Nichols, 2003),no effect for FMRFamide in triggering eclosion has been uncovered(Fuse and Truman, 2002; Zitnan et al., 2003). In fact, FaRPs accumulate in a model of developmental arrest in Manduca(Zitnan et al., 1995). Even if overexpression of FMRFa in all neurosecretory cells in the wit rescue experiment could be due to a novel role for FMRFamide in ecdysis regulation in the CNS, it seems unlikely that this would apply in the wild-type condition, in which the Wit pathway is only regulating systemic FMRFamide. For the systemic peptide to have any central effect it would have to cross the blood-brain barrier that sheaths the nervous system(Carlson et al., 2000; Kretzschmar and Pflugfelder,2002). Although this is possible, it seems more likely that the systemic delivery system has specifically evolved to function only on peripheral tissues and not the CNS itself. Effects of FMRFs in the CNS seem more likely to be mediated by the other neuroendocrine cells that are not affected in wit mutants. It is apparent that a complete dissection of the roles of systemic FMRFamide verse central FMRFs in synaptic transmission and eclosion will need to await the isolation of FMRFa mutants.

How Wit signaling regulates FMRFa expression is not clear. As Smads are well known to act as transcriptional co-activators or co-repressors(Attisano and Wrana, 2002; Moustakas et al., 2001) the simplest explanation is that Mad directly regulates activation of FMRFa transcription, perhaps by forming a complex with Ap. However,other indirect mechanisms are also possible and this issue will only be resolved once the FMRFa promoter is fully characterized. It is also not clear whether Gbb is the only ligand that regulates FMRFaexpression through Wit. In some developmental contexts, such as wing imaginal disc patterning, Gbb acts in combination with Dpp, another Bmp-type ligand(Haerry et al., 1998; Khalsa et al., 1998). We do not see any expression of dpp in the NHO. However, it could be that one of the as yet uncharacterized ligands, Maverick or Myoglianin, could be a partner with Gbb in regulating FMRFa expression. Conversely, it seems clear that regulating the peptidergic phenotype of the six Tv neurons is not the only role of Gbb signaling. There are hundreds of neurons that receive Bmp signaling as indicated by P-Mad nuclear localization Most of them appear to be motoneurons, which require Wit/Gbb signaling to achieve proper synaptic growth but not to specify their neurotransmitter phenotype (Aberle, 2002;Marqués, 2002). Given that Smads act as co-transcriptional regulators,the fact that the same signal (nuclear translocation of P-Mad) results in different phenotypic outcomes in different neurons can probably be ascribed to the presence of a different set of transcription factors available in each cell type. The Tv neurons receiving the Bmp signal express apterous,a transcription factor required in those cells for FMRFatranscription (Benveniste, 1998), and maybe other factors that are required,in addition to the Wit signal, to activate FMRFa.

Another important issue to resolve is whether Gbb is constitutively released from the NHO, or is synthesized and released as part of a feedback mechanism to modulate muscle contractions. It might be that efficient muscle contraction under normal conditions requires a constant level of FaRPs that are produced in response to a constitutive Gbb signal. Alternatively, Gbb production or release might be regulated by a sensing mechanism that would activate the pathway in response to an increased demand for FaRPs, owing to increased locomotor activity or other stimuli, such as compensating for a synaptic developmental defect. We have recently found that muscle-derived Gbb acts through neuronal Wit to convey a retrograde signal essential for NMJ synapse growth and maturation (McCabe et al., 2003). In that context, it appears that the role of Bmp signaling is to coordinate muscle growth with synapse maturation to ensure proper synaptic efficiency. Thus, our combined results indicate that the Wit/Gbb pathway acts as a two-step regulator of NMJ function. First, there is a developmental role in which Wit signaling is required for proper synaptic growth during larval development (Aberle et al., 2002; Marqués, 2002; McCabe et al., 2003). Second,Wit signaling is required to achieve the neuromodulatory effect of circulating FaRPs that are required for optimal synaptic transmission. Lack of either one of these inputs probably results in a substantial decrease of the EJCs. These two examples suggest that the Gbb/Wit pathway is of general importance in neural retrograde signaling and we speculate that it may be used in the nervous system for other as yet uncharacterized developmental and physiological purposes.

Tv neurosecretory cells form part of a cluster of four apterous-expressing neurons on each side of the three thoracic ganglia. The axons of the Tv neurons extend proximally and dorsally to join the contralateral axon, and form a median nerve that swells and arborizes onto a group of neurons and glial cells that constitute the neurohemal organ(Benveniste et al., 1998). In wit mutants, these structures develop normally, but the Tv neuron fail to activate FMRFa transcription. Using the Gal4/UAS system we narrowed Wit's requirement for FMRFa expression to the Tv neurons. As these neurons accumulate nuclear P-Mad, the results strongly suggest that Wit is required in the Tv neurons themselves, as opposed to forming part of an indirect signal relay mechanism. It appears likely that the source of Gbb in this signaling system is the NHO, as gbb is expressed in the NHO and replenishing Gbb in the NHO of gbb mutants rescues FMRFaexpression in the Tv neurons. These experiments do not exclude the possibility that signaling might occur at the cell soma of the Tv neurons in vivo or that the source of the diffusible ligand could be a different tissue under physiological conditions. However, the dependence of nuclear P-Mad accumulation and FMRFa expression in Tv neurons on Dynein-mediated retrograde transport strongly suggests that signaling is taking place at the Tv axon terminal. This dependency on Dynein motors is not a general requirement for FMRFa expression in all neurons because subesophageal ganglion neurons are not affected by overexpression of dominant-negative Glued or Dynamitin. Nor is the consequence of disrupting this motor likely to exert its effect at the level of P-Mad translocation to the nucleus, as nuclear accumulation of P-Mad in epithelial and mesodermal cells is not effected by retrograde transport disruption. Only in the nervous system is P-Mad accumulation specifically affected (McCabe et al., 2003), consistent with a role for a retrograde transport mechanism in moving some component of this signaling pathway from the synapse to the nucleus.

It is interesting to note that others have recently demonstrated that misexpressing ΔGl in the nervous system results in a decrease of quantal content and the number of synaptic boutons at the larval NMJ(Eaton et al., 2002). These phenotypes are strikingly similar to wit mutants, and are consistent with the notion that disruption of retrograde transport prevents Bmp signals from reaching the nucleus. These investigators suggested that the role of Dynactin is to maintain the presynaptic microtubule cytoskeleton thus contributing to synapse stability. However, our data and experiments described elsewhere (McCabe et al.,2003) indicate that failure to traffic a Gbb signal from axon to nucleus also likely contributes to the synaptic defects exhibited by overexpression of ΔGl in motoneurons.

The block of P-Mad accumulation and FMRFa expression upon disruption of retrograde transport suggests that one of the components of the signaling pathway has to be transported along the microtubules from the axon arborizations to the neuron cell body. The normal nuclear accumulation of P-Mad in epithelial and mesodermal cells in which Dynactin function has been disrupted makes it unlikely that the signaling block is at the level of receptor internalization (Penheiter et al., 2002) or translocation of P-Mad from cytoplasm to the nucleus. One possibility is that Mad itself could be phosphorylated at the axon terminal and then transported, either alone or in conjunction with other components, to the nucleus. However, a second possibility is that an activated receptor complex is transported back to the cell body. In the case of Neurotrophins, which constitute the best studied retrograde signaling pathway,it appears that ligand/activated receptor complexes are internalized through a clathrin mediated mechanism, and this signaling endosome is then routed to its appropriate cellular compartment through Dynein mediated retrograde traffic(Ginty and Segal, 2002; Miller and Kaplan, 2001; Miller and Kaplan, 2002; Yano et al., 2001). In the case of Wit, the receptor complex in the signaling endosome would probably contain the activated type I receptors Tkv and Sax(Hayes et al., 2002), and perhaps also Gbb. Mad phosphorylation could occur in the cytoplasm after retrograde transport of the signaling endosome. The dependence of Bmp signaling in the CNS on receptor internalization is supported by analysis of mutations in the spinster gene. Spinster is a component of the late endosomes/lysosomes, and mutations in this gene result in enhanced Wit signaling and synaptic overgrowth, perhaps because of faulty downregulation of signaling endosomes (Sanyal and Ramaswami,2002; Sweeney and Davis,2002).

Proper synaptic transmission requires formation, maintenance and pruning or strengthening of specific synapses in response to developmental or environmental stimuli. Our observations that Bmps control synaptic activity by regulating synapse growth as well as the expression of the neuromodulatory FaRPs, is particularly intriguing in light of another recent report that the Drosophila Activin pathway also regulates synapse function by stimulating pruning of mushroom body synapses during the larval to pupal transition (Zheng et al.,2003). Whether this involves a retrograde Activin signal is not clear, but it emphasizes the notion that this family of growth factors appears to have been recruited multiple times during evolution to regulate different aspects of synaptic function in invertebrates. Consistent with this view are previous reports that in Aplysia, TGFβ can induce long-term synaptic facilitation (Chin et al.,1999; Zhang et al.,1997). Given that various TGFβ-type ligands and their receptors are expressed in specific regions of the adult rodent brain(Charytoniuk et al., 2000), it will be interesting to determine if these proteins also participate in modulating synaptic function in mammals, particularly in regard to long-term memory and learning.

We thank Peter ten Dijke for generously supplying us with anti-pMad antibodies and K. Zinsmaier for anti-Csp antibody. We thank Kristy Wharton for supplying us with gbb alleles, Rahul Warrior for UAS-dynamitin constructs, and Brian McCabe for the G14>Gal4 muscle specific driver. We are indebted to Paul Taghert for supplying the C929 and 386 Gal4 lines and for many helpful discussions. We thank Hong Bao for initial contributions to the electrophysiological project. We acknowledge Tom Hays for suggesting the experiments of retrograde transport disruption. This manuscript was improved by the thoughtful comments of Mary Jane Shimell, Rahul Warrior and Scott Selleck. B.Z. and M.X. were supported by UT start up funds and a CAREER award(IBN-0093170) from the NSF. M.B.O. is an Investigator with the Howard Hughes Medical Institute.