TGF-α ligands can substitute for the neuregulin Vein in Drosophila development

Drosophila Egfr is the sole homolog of the vertebrate family that includes Egfr/ErbB1, neu/ErbB2, ErbB3 and ErbB4. Perturbation of the pathways, which have key roles in proliferation and differentiation during development and homeostasis, results in major developmental defects and problems in human health, including heart disease, schizophrenia and cancer [reviewed by, for example, Mei and Xiong (2008); Sanchez-Soria and Camenisch (2010); Yarden and Pines (2012)]. In mammals there are 11 EGF-like ligand genes (Groenen et al., 1994; Yarden, 2001). Drosophila has four EGF ligands, representing two major classes: the TGF-α ligands spitz (spi), gurken (grk) and Keren (Krn), and the neuregulin-like (NRG) ligand vein (vn) (Rutledge et al., 1992; Neuman-Silberberg and Schüpbach, 1993; Schnepp et al., 1996; Reich and Shilo, 2002; Urban et al., 2002).

Deployment of the four ligands, which operate both in distinct and in overlapping processes, contributes significantly to Drosophila development (Shilo, 2003). grk has a maternal role (Schüpbach, 1987), whereas spi has a major role in zygotic development and mutants die as embryos with ventral defects (Mayer and Nüsslein-Volhard, 1988). vn mutants have milder ventral defects and die at the pupal stage (Simcox, 1997). spi; vn double mutants have more than an additive phenotype, consistent with redundant ligand function (Schnepp et al., 1996). Further redundancy is suggested by the finding that Krn single mutants are viable and fertile adults, but functions in the eye, gut, ovary and brain are revealed when other ligands are compromised, animal physiology is altered or when Krn is expressed ectopically (Yang and Baker, 2003; McDonald et al., 2006; Brown et al., 2007; Jiang and Edgar, 2012; Rahn et al., 2013). Ectopic Krn expression can restore MAPK signaling in a spi mutant, thereby further demonstrating functional redundancy of the TGF-α ligands (Reich and Shilo, 2002). Some roles, by contrast, are ligand specific; for example, vn has a major role in wing development, whereas spi or Krn are not required (Simcox et al., 1987; Simcox, 1997; McDonald et al., 2006).

Here, we investigated the requirement for ligands by examining mutants lacking combinations of genes, and for ligand specificity by determining the ability of one ligand to replace another. We found that Egfr signaling is fully ligand dependent in embryogenesis and that all zygotic ligands play a role in embryo and wing development. Surprisingly, each of the TGF-α ligands could largely replace the essential role of vn in development when the active ligand was expressed in the vn pattern. Our results support the idea that stimulating the pathway to a particular level, and not the specific ligand bound, is important.

As single mutants, only spi and vn have defects in embryonic ventral patterning, and the double mutant is more severe (Schnepp et al., 1996) (Fig. 1A-C). Here, we found that Krn plays a redundant role with spi in embryonic cuticle patterning. spi; Krn double mutants had a similar phenotype to spi; vn double mutants (Fig. 1C,D). The triple mutant spi; vn; Krn was much more extreme than either double mutant and was indistinguishable from an Egfr null mutant (Fig. 1E,F). Both mutants lack ventral denticle belts and head structures. As expected, zygotic removal of the maternal ligand grk (spi, grk; vn, Krn) had no additional effect (Fig. 1G).

The equivalent phenotypes of mutant embryos lacking the receptor or the ligands suggest that there is no signaling through Egfr in the absence of a ligand (Fig. 1E-G). Ligand-independent signaling has been reported for overexpression of Egfr in Drosophila and in mammalian cells, where the high levels of Egfr are thought to cause spontaneous dimerization and signaling (Schweitzer et al., 1995; Nagy et al., 2010; Endres et al., 2013). Our results suggest that, at the normal physiological level of the receptor in the fly embryo, all signaling is ligand dependent. The mutant combinations, however, suggest extensive redundancy between the ligands, consistent with roles that are only revealed when multiple ligands are compromised.

vn has a role in patterning of the wing blade into territories of veins and interveins (Sturtevant and Bier, 1995; Simcox et al., 1996). No role has been found for spi or Krn in wing patterning (Simcox, 1997; Guichard et al., 1999; Nagaraj et al., 1999; McDonald et al., 2006). Reducing activity of Egfr or the protease Rhomboid (Rho), which is required to process the TGF-α ligands, causes vein loss, and rho vn double mutants lack all veins (Sturtevant et al., 1993; Sturtevant and Bier, 1995; Guichard et al., 2000; Ghiglione et al., 2002; Urban et al., 2002). This suggests that either or both spi and Krn could have redundant roles in wing development. We found that eliminating Krn function exacerbated the vein-loss phenotypes in vn mutants (Fig. 1H-K). Reducing spi levels had no observable effect on vn phenotypes (Fig. 1L; supplementary material Fig. S1). The most extreme phenotype resulted from reduction of function in all three genes (Fig. 1M), thereby suggesting that all three contribute to wing vein patterning.

The vn-GAL4 allele, generated as part of this study, has an amorphic phenotype and mutants die as white pupae with tiny wing disks (Fig. 2D-F; supplementary material Fig. S2). This allele, which is expressed in the endogenous vn pattern (supplementary material Fig. S3), allowed us to test whether ectopic activity of the TGF-α ligands rescued vn mutants.

Expression of UAS-vn with vn-Gal4 (UAS-vn; vn^L6/vn-Gal4) rescued ∼30% of vn mutants to adult flies with grossly normal patterning of the wing (Fig. 2G-I). Some extra vein material was observed, consistent with mild Egfr overactivity (Fig. 2I). Expression of rho (UAS-rho; vn^L6/vn-Gal4) rescued ∼50% of vn mutants to pharate adults (differentiated adults that remain in the pupal case) with grossly normal wings (Fig. 3K,L). The region between veins L3 and L4 was fused and there was extra vein material, which correlates with the stripe of vn expression in the wing pouch (Simcox et al., 1996), suggesting that vn is uniquely suited to pattern this region correctly. vn is also required during the first and second instars for growth and global patterning of the early wing disk (Wang et al., 2000; Paul et al., 2013). Expression of rho replaced this function of vn and produced a disk of normal size and pattern (Fig. 2K). This suggests that one or more TGF-α ligands are present in the disk at the early stages, but remain in an inactive form because the processing factor rho is not normally expressed (Simcox, 1997; Zecca and Struhl, 2002).

Rho can cleave all three TGF-α ligands, and to determine whether expression of a particular ligand can replace the function of vn, we tested each ligand individually. Secreted forms were expressed to bypass the requirement for Rho (Urban et al., 2002). Expression of any TGF-α ligand rescued development of vn mutants to the pharate adult stage (Fig. 3; supplementary material Figs S4 and S5). The wings had regions of vein fusion and extra-vein material, characteristic of Egfr overactivity (Fig. 3).

The level of TGF-α transgene expression had a significant impact on the extent of rescue (supplementary material Fig. S4). To correlate rescue phenotypes with signaling, we examined in more detail the ability of secreted Spi (sSpi) to rescue vn mutants. Expression of sSpi at varying levels (supplementary material Fig. S6) produced different degrees of rescue in vn mutants (Fig. 4A). Adult head, leg and abdomen development were restored across a range of expression levels tested, suggesting that these body parts are relatively insensitive to the level of Egfr activity (Fig. 4A; supplementary material Figs S4 and S5). By contrast, a progression of rescue phenotypes was seen, as sSpi expression increased, from animals lacking all wing disk structures, to those with just wings and to optimal cases of animals with wings and a thoracic body wall (Fig. 4A I-III; supplementary material Fig. S7). At the highest Spi levels tested, the animals died as larvae (Fig. 4A IV). The Spi expression level that gave rise to optimal rescue correlated with a signaling output, as measured by MAPK phosphorylation (dpErk), which most closely matched to wild type (WT) (Fig. 4B,C). High levels of sSpi expression produced strongly elevated dpErk levels, which most likely accounted for the early death of these animals.

The intrinsic activity of an EGF ligand correlates with its biological effect; thus, at saturating concentrations, a ligand with high intrinsic activity has a higher maximum biologic effect than a ligand with low intrinsic activity (Wilson et al., 2012). Spi, therefore, appears to have a higher intrinsic activity than Vn, because at higher levels Spi exceeded the required biologic response and was toxic (Fig. 4; supplementary material Fig. S4). Spi also has a higher affinity than Vn (Alvarado et al., 2010). Low affinity might limit the access of Vn to the receptor through negative cooperativity, which proposes that, when the first ligand binds, a conformational change occurs that occludes the second site, rendering it accessible only to a high affinity ligand (Macdonald and Pike, 2008; Alvarado et al., 2010). The biological role of a low-affinity ligand like Vn might therefore be limited, because some cell responses require signaling levels that can be evoked only by high levels of a high-affinity ligand (Krall et al., 2011).

Even in cases of the best morphological phenotypes, rescue of vn mutants by spi was incomplete because the flies did not eclose. This could reflect the artificial Gal4-UAS expression system. Alternatively, the Ig-domain found in Vn and the vertebrate NRGs could confer a unique function. In mice, Ig-containing NRG isoforms are essential (Kramer et al., 1996). A form of Vn lacking the Ig-domain (Vn-ΔIg) rescued vn mutants to adults, thus demonstrating that Ig containing forms are not essential for viability (supplementary material Fig. S8A). The wings of these flies were abnormal and had notched margins (supplementary material Fig. S8B), a phenotype that was also seen when this transgene was expressed in a wild-type fly (Donaldson et al., 2004). The Ig-domain could mediate the role of Vn in another pathway independent of its role in Egfr signaling. The wing phenotype is reminiscent of Notch or wingless mutants (Couso et al., 1994; Price et al., 1997; Wang et al., 2000), and the Vn Ig-domain physically interacts with Hedgehog co-receptors (Özkan et al., 2013), making any of these pathways candidates for interacting with Vn via the Ig domain.

We found a chimeric ligand, in which the Spi EGF-domain had been swapped with the Vn EGF-domain (Schnepp et al., 1998), rescued a vn mutant. These animals developed to pharate adults with pattern defects consistent with Egfr overactivity (supplementary material Fig. S8C,D). The phenotypes were similar to those seen following rescue by sSpi, showing that linking the Spi EGF-domain with an Ig-domain does not significantly alter its activity relative to the receptor.

In summary, our results show a high degree of functional redundancy among Egfr ligands and, moreover, that low levels of a strong TGF-α ligand like Spi can largely replace the developmental function of a weaker ligand like Vn. This is not reciprocal, however, because the lower biological activity of Vn limits its ability to replace or emulate the activity of Spi (Schnepp et al., 1998; Golembo et al., 1999; Donaldson et al., 2004; McDonald et al., 2006; Jiang and Edgar, 2009). The results support the idea that the appropriate cell response will occur regardless of ligand identity, provided a certain quantitative level of signaling is met. This conclusion is based on signaling resulting from activity of the EGF domain, which all Egfr ligands possess. Other unique domains, such as the Ig-domain found in Vn and the neuregulins, might facilitate engagement of other pathways and hence regulate different targets.

In the first example documenting distinct responses to different growth factors engaging MAPK signaling, the duration of signaling was shown to determine whether cells differentiated or proliferated (Marshall, 1995). The subsequent understanding of pathway architecture, including feedback loops, provides a deeper mechanistic view of how intracellular signaling through shared pathways produces different transcriptional responses (Nakakuki et al., 2010). Determining the amplitude and duration of signaling and specific transcriptional outputs induced by Spi and Vn will help shed light on potential differences between the ligands. These might account for the incomplete rescue of vn mutants by the TGF-α ligands and the patterning defects they induced in the wing. Nevertheless, the substantial rescue resulting from ligand substitution prompts the question of why multiple ligands have evolved in the fly when it seems that, in theory, fewer could suffice. Complex animals like the fly might have evolved a collection of high- and low-affinity ligands that are active in specific patterns and at different levels because this is more parsimonious than deploying a multitude of tissue-specific modulators to fine-tune the actions of fewer ligands.

The following mutant alleles were used: spi^2A, grk^HF, vn¹, vn^L6, Krn²⁷, Krn⁹, Egfr^3F18 and Egfr^F24. The following transgenes were used: vn-Gal4, sd-Gal4, UAS-vn1.1 (Schnepp et al., 1996); UAS-vnΔIg (Donaldson et al., 2004); UAS-rho, UAS-mspi::GFP (Tsruya et al., 2002); UAS-sspi (Schweitzer et al., 1995); UAS-sKrn (Urban et al., 2002); UAS-sgrk (Queenan et al., 1999); UAS-vn::SpiEGF (Schnepp et al., 1998); UAS-spi^dsRNA (TRiP 34645 and 28387); UAS-vn^dsRNA (VDRC 50358 and 109437) and tubP-Gal80^ts. Stocks with combinations of alleles were generated from single mutants, with the exception of a spi^2A, grk^HF/CyO stock that was a gift from Trudy Schüpbach (Princeton University, NJ, USA). The vn-Gal4 allele was obtained using the method of Sepp and Auld (1999). The vn-lacZ element (P[PZ, ry⁺]vn^rF264, FBti0005059) was replaced with the P[GawB, w⁺] element in the PG142/FM7 stock (a gift from Norbert Perrimon, Harvard Medical School, Boston, MA, USA). The Gal4 element in the vn-Gal4 allele is in the 5′-UTR, 590 bp upstream of the open reading frame. Krn recombinants were genotyped by PCR, as homozygotes have no phenotype. Primers were used to amplify across the deleted region (8 kb in WT) and gave a band of ∼1 kb for Krn²⁷ or ∼1.2 kb for Krn⁹ (F: AGTCGGCGGCCGTCAATTCG, R: TCCTGGGGCTCCT-TGCGTGT).

Immunoblotting and immunohistochemistry were conducted using standard protocols with the following antibodies: anti-Phospho-p44/p42 MAPK (Cell Signaling Technology, #9106; 1:1000), anti-ERK 2 (Santa Cruz Biotechnology, K-23; 1:2000), anti-Tshirt (a gift from Steven Kerridge, Developmental Biology Institute of Marseille, France; 1:500), anti-GFP (Sigma, #G 1544; 1:1000) and anti-Wingless (Developmental Studies Hybridoma Bank, 4D4; 1:166).

We thank our colleagues Henri Jasper, Denise Montell, Norbert Perrimon, Trudi Schüpbach and Jessica Treisman, the Bloomington Stock Center for fly stocks, and Cassandra Jones and Adeline Ding, high school students at the time, for technical help.