The Drosophila Ret gene functions in the stomatogastric nervous system with the Maverick TGFβ ligand and the Gfrl co-receptor

The RET (rearranged during transfection) receptor tyrosine kinase is the leading susceptibility locus for Hirschsprung's disease (HSCR), a congenital lack of neurons in the distal regions of the digestive tract (McKeown et al., 2013; Romeo et al., 1994). HSCR arises due to the abnormal migration and survival of enteric neuron precursors derived from the neural crest, which has been classified as a neurocristopathy (Zhang et al., 2014a). RET is also found to have a role in kidney development and in a subset of neuroendocrine cancers (Davis et al., 2014; Romei et al., 2016; Schuchardt et al., 1994). The ligands for RET are members of the Glial cell line-derived neurotrophic factor (GDNF) family, which act by binding to a GDNF family receptor (GFR) to activate intracellular RET signaling, or the Neural cell adhesion molecule (NCAM) (Jing et al., 1996; Paratcha et al., 2003; Treanor et al., 1996). GDNF is an important component of vertebrate brain development and maintenance, with clinical relevance to Parkinson's disease (Ibáñez and Andressoo, 2017).

GDNF ligands appeared with the emergence of jawed fish and GFRs underwent a gene expansion at the same time (Airaksinen et al., 2006; Hätinen et al., 2007; Kallijärvi et al., 2012). This expansion coincides with the appearance of the neural crest, a distinguishing structure for vertebrates. Homologs of the RET and GFR receptors are present in invertebrates but are thought to function independently of each other, with GFRs operating in conjunction with Fas2/NCAM rather than with a soluble ligand (Kallijärvi et al., 2012) and RET operating with integrins (Soba et al., 2015). In Drosophila, the RET gene (Ret) is expressed by enteric neurons and epithelial progenitor cells of the adult midgut and is required for homeostasis of these populations (Perea et al., 2017). In the Drosophila embryo, Ret is expressed in the developing stomatogastric nervous system (SNS), a population of cells that delaminate and migrate along the developing gut to form the enteric nervous system (ENS), and Ret is also expressed in the Malpighian tubules, the fly equivalent of the kidney (Hahn and Bishop, 2001; Copenhaver, 2007; Hartenstein, 1997). We previously observed expression of Gfrl promoter fragments in the developing SNS, suggesting that Ret and Gfrl might function together in this tissue (Hernández et al., 2015). Here, using CRISPR, we generated Drosophila Ret alleles and found defects in embryonic SNS formation and larval SNS function. These phenotypes led us to identify the novel TGFβ family member Maverick (Mav) as an invertebrate GFR/Ret ligand and a candidate for the ancestor of GDNF. Our results reveal remarkable similarities in the signaling mechanisms used to generate the insect SNS and the vertebrate ENS.

The role of the Drosophila Ret gene in dendrites has previously been analyzed using a transposon insertion in the 3′ UTR of Ret and the adjacent Mcm10 gene (Soba et al., 2015). We sought to generate Ret alleles that disrupted the coding region of the gene using the CRISPR/Cas9 system. We designed guide RNAs (gRNAs) to a site immediately downstream of the signal sequence and also to the cadherin-like domain (CLD). The gRNAs were introduced as transgenic constructs and crossed to sources of Cas9 (Port et al., 2014). The frequency of induced mutations was >90% and three alleles were selected for further analysis: two separate deletions that lead to truncated proteins (LM1, LM3), and an in-frame deletion that removes a tyrosine conserved with the human RET protein (LM2) (Fig. 1A).

We generated an antibody against the Ret ectodomain that recognizes a band of ∼150 kDa when Ret is expressed in cell culture. Immunoblot analysis of embryonic extracts reveals an absence of this band in all three Ret alleles (Fig. 1B), suggesting they are likely to be null alleles. Homozygous larvae were found to hatch and display a foraging phenotype previously seen in mutants with feeding defects (Fig. 1C) (de Belle et al., 1989; Zinke et al., 1999). To evaluate this phenotype, we added Carmine Red dye to yeast paste to examine feeding behavior (Melcher and Pankratz, 2005), and found Ret mutant larvae frequently had food stuck in their esophagus (Fig. 1D). In addition to this striking phenotype, foraging larvae were often found with reduced or absent food in their midguts and without food in their esophagus. These larvae were frequently immobile or sluggish in response to touch. We tested combinations of the three Ret alleles and observed feeding defects in all of them (LM1/LM2, n=55; LM1/LM3, n=40; LM2/LM3, n=46 homozygous larvae examined; Fig. S1). Ret mutant larvae displayed an eating defect 2-4 days after hatching, and failed to grow at a discernible rate (Fig. 1E).

In parallel, we tested a null allele of Gfrl, the fly GFR co-receptor homolog, and found the same phenotype (Fig. 1D). The Gfrl mutant larvae displayed the same range of feeding defects as Ret mutants, with food levels frequently reduced or absent in larvae displaying foraging and sluggish phenotypes. When larvae were separated from their heterozygous siblings on grape agar plates, 82% made it to the pupal stage (n=109), indicating a mortality rate of 18%. Under crowded conditions on grape agar plates in which both heterozygotes and homozygotes were present, homozygous Gfrl larvae displayed a developmental delay of >2 days and decreased viability (3 homozygotes out of a total of 66 larvae scored after 2 days). This suggests that the homozygotes are at a competitive disadvantage. The similarity of the feeding defects suggests that Gfrl functions with Ret for normal larval feeding, in contrast to the proposed independence of GFR in other tissues (Kallijärvi et al., 2012).

During characterization of the CRISPR Ret alleles, a recessive lethal mutation mapping 22 cM from the Ret locus was detected and removed by recombination. This allowed homozygous adults to be recovered. Larvae still displayed eating defects (Fig. 1F) and significant rates of larval mortality (Fig. 1G). Previous work had shown that Ret is strongly expressed in the developing SNS (Hahn and Bishop, 2001), and our phenotypic findings are consistent with a developmental role in this system. Similarly, the expression of Gfrl promoter fragments in the embryonic SNS (Hernández et al., 2015) and the observed feeding defects suggest that Gfrl could function with Ret in the SNS.

To determine the origins of the larval feeding defect, we first examined the development of the embryonic SNS. Drosophila Ret is expressed in three migrating cell clusters that give rise to the SNS (Hahn and Bishop, 2001). The principal features of the embryonic SNS are the frontal ganglion, which lies anterior to the brain commissure; the frontal nerve, which projects anteriorly from the frontal ganglion along the dorsal surface of the pharynx; the recurrent nerve that projects posteriorly to the two esophageal ganglia; and the ventricular ganglion (Fig. 2A,E,I).

We used the w¹¹¹⁸ stock as a control for all experiments (Fig. 2E-E″). Expression of a transgenic RNA interference (RNAi) hairpin for Ret in the developing SNS frequently led to missing or shortened frontal nerves and disrupted frontal ganglia (Fig. 2B,C,F), and to esophageal ganglia that had not migrated as far as in the control (Fig. 2F′). The same defects were seen in the CRISPR-induced Ret mutants (Fig. 2G,G′), confirming a role for Ret in embryonic SNS development.

To prove that the defects were due to a loss of Ret activity, we expressed a Ret transgene in the developing SNS of Ret^LM1 mutants and observed rescue of the frontal and recurrent nerve phenotypes, but not those of the frontal or esophageal ganglia (Fig. 2H,H′). We scored the defects shown in Fig. 2B-D,J and quantified them (Fig. 3A-D), confirming that the Ret CRISPR mutants and Ret RNAi have very similar phenotypic defects. None of the phenotypes documented is 100% penetrant, so variability in larval feeding might be due in part to variation in embryonic development. Among the foraging first instar larvae, we observed midgut axon morphologies strikingly similar to those of older animals (Fig. S2A). Differences between alleles might be due to background effects of particular chromosomes or to natural variability in the structures examined. Similarly, for the rescue experiments, lack of statistically significant rescue in the frontal or esophageal ganglia might be due to the variability of these structures or the promoter used might lack adequate expression. In adult flies, Ret function is required in epithelial cells of the intestine (Perea et al., 2017). The Ret-GAL4 lines used express in the epithelial lining of the embryonic midgut, particularly in the earlier stages (Hernández et al., 2015), so we cannot rule out Ret function in the gut as a cause for the SNS patterning defects. However, we used the lines (P2A and P2B) that had the lowest gut expression and that display prominent SNS neuron expression. Gut expression is minimal in the later stages of embryogenesis and undetectable in the first instar larva (Fig. S2B). Overall, the results indicate a role of Ret in the normal patterning of the embryonic SNS.

The relatively subtle nature of the embryonic SNS defects in Ret mutants led us to examine the anatomy of the larval SNS. Second and third instar (2-5 days after hatching at 25°C) Ret mutants had food stuck in their esophagus immediately anterior to the proventriculus, a food-grinding organ and valve (Fig. 1D, Fig. 4A,B). The proventriculus is an infolding of the gut at the junction of the foregut and midgut that regulates food entry into the midgut (Spieß et al., 2008). The proventricular ganglion constitutes the posteriormost cell bodies of the SNS and is located at the anterior end of the proventriculus. The proventricular ganglion forms the midgut nerves, with axons projecting to the anteriormost end of the midgut (Gonzalez-Gaitan and Jackle, 1995).

We examined the midgut nerves of second and third instar larvae with feeding defects by staining for Futsch (mAb 22c10) (Fig. 4). Axons normally project from the proventriculus (ventricular ganglion) neurons posteriorly in three to five well-defined fascicles (Fig. 4C) (Spieß et al., 2008) and begin branching upon encountering the midgut (Fig. 4C′). In Ret mutants, the axon bundles often defasciculate a short distance into their trajectories (Fig. 4D). The most obvious phenotype in Ret mutants is increased axon branching upon encountering the midgut, although the axons cover a smaller area than in w¹¹¹⁸ (Fig. 4D′). We quantified this phenotype using ImageJ and the Simple Neurite Tracer plugin (Longair et al., 2011) and found that Ret mutants showed a 2.25-fold increase in the frequency of axon branching that is highly statistically significant [Fig. 5A; P<0.001, Tukey honestly significant difference (HSD) test]. To confirm that the defect was due to lack of Ret, we expressed a Ret transgene in the SNS neurons and found that the proventricular defasciculation (Fig. 4E) and the midgut axon branching phenotypes (Fig. 4E′) were no longer present.

As noted above, the Ret-GAL4 lines express predominantly in SNS neurons, and this pattern of Ret expression continues into the second instar (Fig. S2C). In late second instar and third instar larvae, strong midgut expression is observed, but this is in the middle of the midgut at a considerable distance from the midgut axons. We therefore expect Ret to play a cell-autonomous role in the midgut axons but cannot rule out functions in other tissues. Quantification of the axon branching confirmed the rescue (Fig. 5A).

The midgut neurons are characterized by varicosities, which are likely to be en passant synapses signaling to the underlying gut tissue (Fig. 4C″) (Budnik et al., 1989; Neckameyer and Bhatt, 2012). The Ret mutants frequently display larger varicosities with decreased 22c10 antigen [a microtubule-associated protein (Hummel et al., 2000)] in the axons between the varicosities (Fig. 4D″). The morphology of the axons is reminiscent of the axon fragmentation observed during Wallerian degeneration. Wallerian degeneration can be blocked by reducing the activity of dSarm (Ect4) (Osterloh et al., 2012). However, reducing the activity of dSarm via RNAi failed to suppress the axon branching phenotype or the enlarged varicosities (Fig. 4F,F′, Fig. 5A). To further understand the SNS defect, we dissected live larval guts and observed peristalsis of the midgut. In w¹¹¹⁸ larvae, contraction of the proventriculus is followed by a wave of peristalsis that propagates along the midgut (Fig. 5B). In Ret mutants, a similar frequency of proventricular contractions was observed (Fig. 5B), but the wave of peristalsis was mostly absent (Fig. 5C). These observations suggest that the altered neuroanatomy of the midgut neurons reflects an inability to properly signal to the midgut muscles and this results in a midgut contraction but no propagation of the peristaltic wave.

Vertebrate RET is known to function as a complex, and the Drosophila GFR homolog Gfrl is required for larval feeding (Fig. 1D). We therefore analyzed the embryonic SNS in Gfrl mutants and found striking phenotypic similarities to Ret mutants, with asymmetry of the frontal ganglion and disruption to the esophageal ganglia (Fig. 6B,B′, compare with Fig. 6A and Fig. 2C). GDNF is a member of the TGFβ superfamily and in the fly there are seven TGFβ family members. We examined the expression patterns of the seven TGFβ genes and aligned the amino acid sequences to vertebrate RET ligands. None of the proteins contains the distinguishing motifs of the vertebrate ligands, but the Maverick (Mav) protein showed the best, albeit weak, sequence similarities (Fig. S3). The mav gene is expressed in the developing esophagus and midgut at the right time to act as a ligand (Nguyen et al., 2000). Two independent chromosomal deficiencies for the mav region, Df(4)C1-7a and Df(4)ED6380, displayed phenotypic similarities to Ret mutants (Fig. 6C,C′) (Sousa-Neves et al., 2005). These results suggested that a Ret-Gfrl-Mav complex might be functioning in fly SNS development.

Ret signals through classical receptor tyrosine kinase pathways and genetically interacts with the Pink1 mitochondrial kinase (Ibáñez, 2013; Klein et al., 2014), which has been proposed as a susceptibility locus for HSCR (Meka et al., 2015). Pink1 mutants displayed asymmetric frontal ganglia and altered esophageal ganglia (Fig. 6D,D′), suggesting that Pink1 is functioning in the same pathway.

mav mRNA is expressed in the foregut and esophagus (Nguyen et al., 2000). We looked for dose-sensitive genetic interactions by examining Ret/+; Gfrl/+ transheterozygotes but did not observe any feeding defects or decreases in viability relative to sibling flies. As both mutations are homozygous viable, albeit with decreased viability, we believe sensitized backgrounds will have to be used to demonstrate an interaction.

To examine Mav protein expression, we stained a fosmid-based GFP-tagged Mav with anti-GFP (Sarov et al., 2016). Mav expression is observed in the roof of the stomodeum (mouth) when SNS precursors are invaginating from the epithelium to form migrating clusters (Fig. 6E). Expression continues and broadens within the mouth/pharynx to encompass the esophagus and the proventriculus (Fig. 6F-G′). Based on the migratory behavior of the SNS clusters, mav is expressed at the right time and place to function as a Ret ligand. Overexpression of mav in the digestive tract has a strong effect on embryonic morphology as a whole, possibly affecting midgut differentiation, but SNS neurons also appear to increase in number (Fig. 6H,H′), suggesting that they respond to the Mav ligand.

To test whether Drosophila has an equivalent to the RET-GFR-GDNF complex, we expressed epitope-tagged versions of Mav, Gfrl and Ret in COS7 cells and attempted to co-immunoprecipitate the proteins. In vertebrates, GDNF does not bind directly to RET but acts through GFR (Treanor et al., 1996), so we tested for binding between Mav and Gfrl. Co-expression followed by precipitation with anti-Gfrl (V5) and probing with anti-Mav (Flag) demonstrated a strong interaction between Mav and Gfrl (Fig. 7A). The presence of Ret did not alter the interaction. Immunoprecipitation with anti-Gfrl (V5) and probing with Ret revealed that Gfrl and Ret closely interact, both in the presence and absence of Mav (Fig. 7B). Although the interaction appears weak, expression levels of Ret in cell culture were always lower than for Mav or Gfrl (Fig. S4). We also observed a low level of Mav dimers even under the reducing conditions of the SDS sample buffer (Fig. S4). By analogy to GDNF, it seems likely that Mav dimerizes before binding to Gfrl, and Gfrl may already be bound to Ret (Fig. 7C).

To verify the results, we performed cell overlay assays in which expression of Mav in COS7 cells was detected by immunohistochemistry. Cells expressing Mav were clearly detected but weak cell surface staining suggesting diffusion into the medium (Fig. 7D″). When Gfrl was co-transfected, a dramatic increase in cell surface levels of Mav was observed and the Mav staining was highly punctate (Fig. 7E″). Colocalization of Mav and Gfrl was also observed, although the proteins have distinct patterns, with Gfrl being more diffuse than the highly punctate Mav (Fig. 7E‴). Expression of Ret alone led to a slight increase in cell surface levels of Mav, with a few puncta present (Fig. 7F″), suggesting a possible weak interaction. Co-transfection of Ret and Gfrl produces Mav binding equivalent to Gfrl alone (Fig. 7G-G′″). The strong alterations in the localization patterns combined with the immunoprecipitation results suggests that Mav, Gfrl and Ret are present in a physiologically relevant complex.

We have described the effects of mutating the Ret gene in Drosophila and uncovered an evolutionarily conserved role in the development of the ENS. The incorrect positioning of SNS cells in the Drosophila embryo resembles hypoganglionic ENS phenotypes seen when RET is mutated in vertebrates (Lake and Heuckeroth, 2013). In HSCR, the most distal nerves of the digestive tract are affected. Likewise, in Ret mutant larvae we find that the most distal nerves of the SNS, located on the midgut, have an altered anatomy and the larvae show defects in food ingestion. The phenotype resembles the neurotrophic effects of decreased serotonin or CNS dopamine signaling during midgut nerve formation, which also leads to increased axon branching and decreased feeding (Budnik et al., 1989; Neckameyer and Bhatt, 2012).

Although defects are visible in the embryonic SNS, there appear to be two separate lethal phases. Some first instar larvae display feeding defects and die. This is particularly evident in the original alleles that carry the background recessive lethal mutation, and we are investigating the possibility that the background lethal mutation specifically enhances the Ret mutations. Subsequent larval feeding defects often do not emerge until 2-4 days after hatching. Larvae with food in their guts can be observed foraging, suggesting that the larvae have problems with food ingestion. This is supported by observations of mutant larvae with food throughout their midguts, but with peristaltic defects in the anterior midgut (see Movie 2, compare with wild-type peristalsis in Movie 1). We initially suspected a neurodegenerative defect similar to Wallerian degeneration, but failed to suppress the axon defect by reducing dSarm activity. We currently favor a model in which initial SNS defects are amplified as the larva dramatically increases its mass several hundred fold (Ghosh et al., 2013). To keep pace with the expanding midgut, Ret may be required to promote axon growth, guidance, or be fulfilling a pro-synaptic role. These functions have been observed for RET and GDNF (Dudanova and Klein, 2013; Paratcha and Ledda, 2008), including in the human ENS (Böttner et al., 2013).

The midgut axon phenotype resembles defasciculation of the nerves and Gfrl genetically interacts with the fasciculation molecule Fas2 (Kallijärvi et al., 2012), so Ret/Gfrl could potentially be modulating fasciculation as has been observed for other signaling systems (Yu et al., 2000). Alternatively, defasciculation may be a consequence of growth cones searching for sources of ligand, as proposed for Netrin and Bolwig's nerve (Andrews et al., 2008). Decreased midgut innervation and function may provide negative feedback to upstream gut signaling, decreasing the ability to pass food through the pharynx and esophagus (Melcher and Pankratz, 2005; Zhang et al., 2014b). The midgut axons may also be required to maintain communication with downstream enteroendocrine cells (LaJeunesse et al., 2010). An alternative hypothesis raised by the similarity of the Ret and Pink1 phenotypes is that the midgut neurons are running out of energy due to mitochondrial dysfunction (Klein et al., 2014).

Our analysis enabled us to identify the divergent TGFβ Mav as the elusive ligand for Drosophila Ret (Kallijärvi et al., 2012; Saarenpää et al., 2017). The expression pattern of mav is consistent with a role in embryonic SNS development (Fig 6E-G, Fig. 8). Although the Mav ligand is concentrated in certain regions of the foregut and may create localized gradients, the broad expression pattern suggests that the Ret/Gfrl signaling pathway could be permissive rather than instructive during SNS precursor migration. Embryonic Ret signaling could primarily transduce a neurotrophic signal, and apoptosis has been observed in the migrating SNS precursors (Hartenstein et al., 1994). In vertebrates, models in which GDNF/Ret signaling promotes proliferation rather than cell migration have been proposed to explain development of the nervous system (Newgreen et al., 2013). Experiments are underway to distinguish between these models in the fly. Although Gfrl expression has not yet been observed in the SNS, Gfrl could be acting in a soluble form or in trans (Paratcha et al., 2001). Gfrl promoter fragments continue to drive expression in the anterior midgut of the larvae in support of the trans model (Hernández et al., 2015). Despite extensive sequence divergence in the extracellular domain of Ret (Hahn and Bishop, 2001), domain differences in GFRs (Airaksinen et al., 2006) and low homology of Mav to the GDNF family (Nguyen et al., 2000), the molecular logic of the protein complex appears preserved. In vertebrates, RET and GFR form a preassembled complex (Eketjall et al., 1999), and GDNF binds GFR to activate RET (Treanor et al., 1996). Our molecular data are strikingly similar, as we find that Drosophila Ret and Gfrl can functionally interact in the absence of Mav, and that Mav interacts strongly with Gfrl, but only very weakly with Ret. In flies, Mav modulates synapse formation at the neuromuscular junction of body wall muscles (Fuentes-Medel et al., 2012; Sulkowski et al., 2016). Ret is not expressed in body wall muscles (Hahn and Bishop, 2001), and Mav is likely to be signaling through activin/BMP type 1 receptors (Fuentes-Medel et al., 2012). A Mav homolog, Panda, has been found in the sea urchin Paracentrotus lividus, where it plays a role in dorsoventral axis formation and is also likely to be signaling through type 1 receptors (Haillot et al., 2015). Mav and Panda both lack a key leucine residue, so their binding to type 1 receptors might be weaker than other ligands (Haillot et al., 2015). Candidate Ret and Mav homologs have been found in Strongylocentrotus purpuratus (Lapraz et al., 2006), suggesting that Mav homologs might interact with both type 1 and Ret receptors in sea urchins.

Despite promiscuity in binding between TGFβ and their receptors in vertebrates, GDNF family members have not been reported to bind BMP/TGFβ receptors (Mueller and Nickel, 2012), suggesting that the ability to interact with more than one receptor was lost during evolution. The GDNF family of ligands, including GDNF, Neurturin, Artemin and Persephin, all appeared when fish gained jaws, as homologs cannot be identified in the published Agnatha sequences. GDNF ligands are distinguished by a highly conserved DLGLGY motif, part of one of two fingers that mediate binding to GFRα (Fig. S3) (Eketjall et al., 1999). This motif is not present in Mav or Panda. The change may have increased affinity or specificity for GFRs and additional changes might have prevented crosstalk with Activin/BMP type 1 receptors. Mav and Panda are similar to GDF-15 (Haillot et al., 2015), a TGFβ placed in the subfamily containing GDNF (Mueller and Nickel, 2012). GDF-15 is an inflammatory cytokine, and although it activates SMAD signaling, GDF-15 does not have an identified receptor (Yadin et al., 2016). GDF-15 has GDNF-like neurotrophic activity for dopaminergic neurons (Strelau et al., 2003), so it would be interesting to test GDF-15 for binding to GFRs.

The limited sequence data available suggest a model in which a divergent TGFβ acquired an ability to bind GFRs and activate Ret, which was followed by extensive co-evolution of the extracellular components. However, the downstream signaling pathways appear to be conserved (Abrescia et al., 2005), so the Ret SNS phenotypes open the door to invertebrate genetic analysis of this clinically important signaling pathway. Particularly exciting is the possibility of functional suppressor screens to identify mutations that could compensate for a lack of Ret signaling. Drosophila has already been used to identify genetic modifiers and a candidate drug to counteract oncogenic Ret signaling (Das and Cagan, 2013; Read et al., 2005).

Ret has an evolutionarily conserved role in the formation and function of the ENS. The GDNF signaling pathway has its origins in TGFβ signaling.

The w¹¹¹⁸ stock was used as a control as it had the most reliable neuroanatomy relative to that described in previous publications. Bloomington Drosophila Stock Center (BDSC) supplied w¹¹¹⁸, Act5C-Cas9, Vasa-Cas9, Df(2L)BSC312[Ret]/CyO, Df(4)ED6380[mav]/l(4)102EFf¹ (Ryder et al., 2007), FasII-GAL4, Pink1⁵/FM6 and Klu⁰⁹⁰³⁶ (stock numbers: #6326, #54950, #52669, #24338, #9579, #46123, #51649 and #11733). BDSC also provides the nos-phiC31 attP2 stock #25710 used for embryo injections by Rainbow Transgenic Flies (Camarillo, CA, USA). The Vienna Drosophila Resource Center (VDRC) provided UAS-RetRNAi and UAS-SarmRNAi, #107658 and #22612 (Dietzl et al., 2007). Balancing of stocks was done using w⁻; Kr^If-1/CyOwgβ, w⁻; Kr^If-1/CyOKrGFP, v⁻; Dr/TM3 and w⁻; UASτlacZ/CyOwgβ; TM2/TM6 (T.K. laboratory stocks). The Ret-GAL4 (RetP2A and RetP2B) lines used in this work were produced and characterized by our laboratory (Hernández et al., 2015). Df(4)C1-7a was supplied by R. Sousa-Neves (Case Western Reserve University). The ΔGfrl1E allele and UAS-Ret^flag were provided by J. Kallijärvi (University of Helsinki). UAS-mav and UAS-mav^GFP lines were provided by V. Budnik (University of Massachusetts Medical School).

A CRISPR strategy to target Drosophila Ret was organized and carried out as described by Port et al. (2014). We received the pCFD3-dU6:3gRNA plasmid courtesy of P. Miura (University of Nevada, Reno; #49410, Addgene). Three individual gRNA target sites within Ret were chosen using the Drosophila CRISPR/Cas9 gRNA target track in the UCSC Genome Browser (https://genome.ucsc.edu) (Speir et al., 2016). Target sites were checked for specificity using the flyCRISPR Optimal Target Finder (http://tools.flycrispr.molbio.wisc.edu/targetFinder/) (Gratz et al., 2014). Single-stranded oligos corresponding to these target sequences were synthesized by Integrated DNA Technologies (San Diego, CA, USA). Sense and antisense oligos for targeting exons 3 and 5 of the Ret locus are listed in Table S1. Three pCFD3-dU6:3-gRNA:Ret constructs (one targeting exon 3 and two targeting exon 5) were produced and screened using PCR primers listed in Table S2. Rainbow Transgenic Flies produced transgenic stocks expressing Ret gRNAs via a phiC31 injection protocol. Recovered flies were balanced with a TM3 balancer and crossed to Act5C/Vas-Cas9 lines. Recovered progeny from the crossing scheme were balanced with CyOwgB or CyoKrGFP. For candidate alleles, the region of interest was amplified by PCR and sequenced. Sequence analysis was conducted using 4 Peaks (https://nucleobytes.com). Three independent lines were isolated after analysis (Ret^LM1-3 alleles). Homozygous embryos for each allele were assayed for protein expression using a polyclonal antibody directed against the Ret ectodomain (residues 24-401; Genscript). Embryos were crushed in 50 µl cell lysis buffer (50 mM Tris-HCl pH 7.6, 1 mM EDTA, 150 mM NaCl, 1% Triton X-100) containing protease inhibitor cocktail (Sigma-Aldrich), incubated on ice for 15 min, cleared by centrifugation and analyzed via SDS-PAGE.

Antibody staining on embryos was performed as described by Patel (1994) with reagents listed in Table S1. 22c10 was used to visualize the developing SNS. Balancer markers and GAL4/UAS expression were visualized with anti-β-galactosidase and anti-GFP antibodies. Typical incubations with primary antibodies were overnight at 4°C and 1 h at room temperature with secondary antibodies. Biotinylated secondary antibodies with VectaStain Elite ABC enhancement were frequently used for stronger labeling of SNS projections in intact embryos. Stage 17 embryos of each genotype were collected at random and scored for the presence, absence or thinning of the frontal nerve, and for defasciculation defects in the recurrent nerve. The arrangement of cell bodies at the frontal ganglion, esophageal ganglion, and proventricular ganglion were also scored. At least 20 embryos were collected for each genotype. Gut preparations of second instar larvae were performed as described (Bhatt and Neckameyer, 2013; Neckameyer and Bhatt, 2012). Larval SNS was visualized using 22c10 and scoring of the SNS in larval preparations was performed in at least ten individual larvae for each genotype. Neural innervation of the midgut was assessed by measuring axon coverage of the innervated midgut tissue and the degree of branching by axons using ImageJ/Simple Neurite Tracer. RapiClear (Cedar Lane Laboratories) was used as a whole-mount medium for imaging late stage 17 embryos and larval gut tissues (Hernández et al., 2015).

Behavioral and developmental effects in larvae were assayed via an egg lay on grape juice agar with access to a yeast food paste mixture containing Carmine powder as a coloring agent (Sigma-Aldrich; 1.5 mg Carmine per 1 g yeast paste) as described (Melcher and Pankratz, 2005; Zinke et al., 1999). Sets of 50 homozygous mutant larvae from overnight egg collections were allowed to develop for a 72 h period at 25°C and monitored at 24 h intervals for feeding and wandering phenotypes under a dissection microscope. Animals were assayed for each genotype to assess mortality rate, approximate growth rates, and observable feeding defects. At least 100 newly hatched first instar larvae per genotype were assayed for mortality and behavioral phenotypes. Movements of the proventriculus and the midgut were observed in semi-intact third instar larvae (Schoofs et al., 2014). The preparation consisted of CNS, SNS, foregut and midgut dissected from exterior cuticle in a bath of Schneider's Drosophila medium at room temperature. Preparations were allowed to equilibrate in medium to allow for spontaneous contractile activity to occur. At least ten animals of each genotype were assessed in independent preparations. Videos of gut motility were recorded using a Leica MZFLIII with a Jenoptik ProgRes C14 plus camera for periods of ∼10 min. Measurement was by counting peristaltic events at the proventriculus and in the midgut over 1.5 min from the initiation of spontaneous contractile activity.

For larval feeding phenotypes, mortality, and embryonic SNS analysis, Fisher's exact test with two tails and 95% confidence intervals was calculated using the GraphPad website (www.graphpad.com/quickcalcs). Statistical significance was tested using Bonferroni correction. For larval proventriculus/midgut contractile activity, a t-test and the Bonferroni correction were performed to determine statistical significance and s.e.m. using the GraphPad website. Statistical analysis of larval midgut neuron axon branching used a one-way ANOVA followed by a Tukey HSD using Statistica (Dell). Sample size was determined using the resource equation method: E=total number of animals−total number of groups. For each experiment a value of E>10 was required. No samples were excluded from the analysis and samples were allocated to experimental groups on the basis of genotype. For the data in Figs 2-5, images or movies of embryos or larvae were given codenames and scored by an experimenter blind to the genotype. All data were tested for fit to a normal distribution. Raw data are available in Table S3.

pcDNA3-FLdRet^cmyc was a gift from C. Abrescia and C. Ibáñez (Abrescia et al., 2005). pcDNA3.1-Mav^cDYK was synthesized from the sequence for the final 112 amino acid residues of the fly Mav protein consisting of the proteolytically cleaved active form of the ligand along with an Ig κ-chain secretion signal added to the N-terminus of the protein for secretion (Genscript). The pSecTag-^V5GfrlA was subcloned from pMT-^V5GfrlA that was a gift from J. Kallijärvi (Kallijärvi et al., 2012). We PCR amplified the N-terminally tagged GfrlA construct from pMT-^V5GfrlA (using forward primer 5′-GGTAAGCCTATCCCTAACCC-3′ and reverse primer 5′-TCATGTCGCCACATCACTC-3′) then subcloned the fragment into pSecTag/FRT/V5-His-TOPO (Invitrogen). Expression of Ret, Gfrl and Mav protein in animal cell lines was performed using COS7 and HEK 293 cells. COS7 or HEK 293 cells at 80% confluence were transfected with DNA expression constructs using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. Ret, Gfrl and mav constructs were transfected individually as well as in combination to assess the ability of each protein to bind in pairs and as a complex.

Immunoprecipitation (IP) assays from cell culture followed a published protocol (Alavi et al., 2016; Banerjee et al., 2010). In brief, HEK 293 cells were transfected, grown for 48 h, lysed in IP buffer (50 mM HEPES pH 7.2, 100 mM NaCl, 1 mM MgCl₂, 1 mM CaCl₂, 1% NP-40) and incubated overnight at 4°C with 1 µg antibody. Proteins were immunoprecipitated with 30 µl GammaBind G Sepharose beads (GE Healthcare) for 2 h at room temperature, run on 4-20% Mini-Protean TGX gels (Bio-Rad), immunoblotted and detected with Clarity western ECL substrate (Bio-Rad) and imaged on a ChemiDoc Touch system (Bio-Rad).

Constructs were transfected into COS7 cells for cell immunofluorescent labeling experiments. Chamber slides were coated with rat tail collagen solution (Thermo Fisher Scientific) diluted 1:4 with PBS and allowed to dry. At 48 h post-transfection, medium was removed from the co-transfected receptor (Ret/Gfrl)-expressing cells and replaced with medium containing Mav. All cells were incubated at 37°C for 1-2 h before rinsing three times in 1× PBS and proceeding with fixation and antibody labeling. Alternatively, all three constructs were co-transfected (1:1) into the COS7 cells. Approximately 48 h post-transfection, medium was removed and the cells were washed with 1× PBS. After rinsing, cells were fixed for 15 min in 4% paraformaldehyde at room temperature. Cells were washed once in cold PBS after fix then once in PBST (PBS with 0.1% Triton X-100). Cells were blocked for 30 min at room temperature with 5% normal goat serum (NGS) in PBST. Primary antibodies were added in 5% NGS in PBST and incubated overnight at 4°C. Cells were washed three times in PBST then incubated for 1 h in fluorescent secondary antibody in 5% NGS in PBST. Secondary antibodies were from the AlexaFluor collection (Table S1Supplementary information). Cells were washed five times in cold PBS following labeling. Cells were mounted in FluorSave containing NucBlue DAPI stain (Millipore) to label nuclei.

We thank P. Miura for advice and reagents for CRISPR; J. Kallijärvi for Gfrl and UAS-Ret reagents; P. Soba for UAS-Ret-mCherry; V. Budnik for mav reagents; C. Ibáñez for Ret reagents; R. Sousa-Neves for Df(4)C1-7a; M. Freeman for Sarm reagents; A. DiAntonio for advice on axon regeneration; G. Hennig for advice on video analysis; S. Ward for advice on the enteric nervous system; L. Heydman and N. Yokdang for help with cell culture experiments; R. Marshall, K. Hernández and other members of the T.K. laboratory for assistance with genetics and imaging; P. Soba and N. Hoyer for communicating results prior to publication; and R. Kellermeyer and T. Gillis for comments on the manuscript. Antibodies were obtained from the Developmental Studies Hybridoma Bank (DHSB) developed under the auspices of the NICHD and maintained by the University of Iowa. Drosophila stocks were obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537). Transgenic fly stocks were obtained from the Vienna Drosophila Resource Center. Funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.