The feedback regulator Nord controls Dpp/BMP signaling via extracellular interaction with Dally in the Drosophila wing

The Drosophila BMP 2/4 homologue Decapentaplegic (Dpp) acts as a morphogen to regulate diverse developmental processes, including wing morphogenesis. Transcriptional feedback regulation of this pathway ensures tightly controlled signaling outputs to generate the precise pattern of the adult wing. Nevertheless, few direct Dpp target genes have been explored and our understanding of feedback regulation remains incomplete. Here, we employ transcriptional profiling following dpp conditional knockout to identify nord, a novel Dpp/BMP feedback regulator. Nord mutants generated by CRISPR/Cas9 mutagenesis produce a smaller wing and display low penetrance venation defects. At the molecular level, nord encodes a heparin-binding protein and we show that its overexpression is sufficient to antagonize Dpp/BMP signaling. Further, we demonstrate that Nord physically and genetically interacts with the Dpp/BMP co-receptor Dally. In sum we propose that Nord acts with Dally to fine tune Dpp/BMP signaling, with implications for both developmental and disease models. Impact statement Functional analyses of the Drosophila homologue of Neuron Derived Neurotrophic Factor reveal a new mode of extracellular heparan sulfate proteoglycan regulation required for proper morphogen action.


Introduction
For over 100 years, secreted factors have been recognized as powerful agents to control tissue growth and spatial patterning during development (Rogers and Schier, 2011). This idea gained particular experimental traction from Spemann and Mangold's organizer transplantation experiment (Spemann and Mangold, 1924). When the dorsal blastopore lip of a weakly pigmented salamander embryo is grafted to the ventral side of a pigmented host, the grafted fragment induces secondary axis formation in neighboring pigmented cells, causing a conjoined twin. Deciphering the chemical basis for this organizer activity, however, remained elusive. In the 1990s, a fuller and more complex picture of cell-cell signaling began to emerge. Molecular analyses of the organizer identified many secreted antagonists of signaling molecules, such as Chordin for Bone morphogenetic protein (BMP), Frizzled-related Protein 1 for Wnt and Cerberus for Nodal (Sasai et al., 1994;Sasai et al., 1995;Bouwmeester et al., 1996;Leyns et al., 1997).
These secreted signaling molecules are repeatedly utilized during development and homeostasis to establish proper cell-cell communication and their misregulation often leads to congenital abnormalities and diseases (Wu et al., 2016;Li and Elowitz, 2019;Zhang and Wang, 2020).
During development, a wide range of signaling molecules act as morphogens to govern diverse processes, such as embryonic dorsoventral (D/V) axis formation and neural and limb patterning in both vertebrates and invertebrates (O'Connor et al., 2006;Affolter and Basler, 2007;Muller et al., 2013;Akiyama and Gibson, 2015b;Bier and De Robertis, 2015). Morphogens provide spatial information to cells in a morphogenetic field by creating a concentration gradient, and direct distinct cell fates in a concentration-dependent manner. Over the last three decades, analysis of the developing Drosophila wing has greatly contributed to our understanding the molecular basis of morphogen action (Lecuit et al., 1996;Nellen et al., 1996;O'Connor et al., 2006;Affolter and Basler, 2007;Akiyama and Gibson, 2015b;Beira and Paro, 2016).
The adult wing develops from a primordium called the wing disc, an epithelial invagination which grows dramatically during larval stages and then everts during metamorphosis to produce the adult structure. During wing development, Wingless (Wg; a Drosophila Wnt), Hedgehog (Hh) and Decapentaplegic (Dpp; a Drosophila BMP2/4) act as morphogens to pattern the wing (O'Connor et al., 2006;Affolter and Basler, 2007;Gradilla and Guerrero, 2013;Akiyama and Gibson, 2015b;Beira and Paro, 2016;Bejsovec, 2018). Among them, Dpp is perhaps the most studied morphogen and exhibits two distinct morphogen actions during this process. In the wing disc, Dpp emanates from a stripe of cells adjacent to the anterior-posterior (A/P) boundary to generate a long-range Dpp morphogen gradient along the A/P axis (Lecuit et al., 1996;Nellen et al., 1996;Entchev et al., 2000;Teleman and Cohen, 2000). Dpp activates downstream signaling through a tetrameric receptor complex consisting of BMP type I and type II receptors (Brummel et al., 1994;Nellen et al., 1994;Penton et al., 1994;Xie et al., 1994;Letsou et al., 1995;Ruberte et al., 1995). In Drosophila, the BMP type I receptors, such as Thickveins (Tkv), phosphorylate the downstream signal transducer Mothers against dpp (Mad), resulting in a phosphorylated Mad (p-Mad) gradient that provides a direct readout of Dpp/BMP signaling activity across the wing disc (Newfeld et al., 1996;Tanimoto et al., 2000). The graded distribution of Dpp/BMP activity in turn induces the nested expression of target genes, including Daughters against dpp (Dad), spalt (sal), optomotor blind (omb) and brinker (brk), to establish the intervein and longitudinal vein regions. During pupal development, Dpp/BMP signaling is further required for wing vein morphogenesis (O'Connor et al., 2006;Blair, 2007;Raftery and Umulis, 2012). The Dpp morphogen, now expressed in the longitudinal veins, not only activates the signaling pathway locally but is also transported into the presumptive crossvein region to shape the morphogen gradient by the actions of extracellular BMP interacting proteins, such as Short gastrulation (Sog, a Drosophila Chordin) (Conley et al., 2000;Ralston and Blair, 2005;Serpe et al., 2005;Shimmi et al., 2005). Dpp/BMP activation in the longitudinal vein and crossvein regions triggers the terminal wing vein differentiation process.
Transcriptional feedback regulation of Dpp/BMP signaling is an essential mechanism for buffering against genetic and environmental perturbations (O'Connor et al., 2006;Lander, 2011;Raftery and Umulis, 2012;Bier and De Robertis, 2015). Previous studies have discovered several Dpp/BMP feedback regulators, such as Division abnormally delayed (Dally), Pentagone (Pent) and Larval translucida (Ltl) (Fujise et al., 2003;Vuilleumier et al., 2010;Szuperak et al., 2011). Dally is a member of the glypican family of heparan sulfate proteoglycans (HSPGs), which consist of a protein core and covalently attached heparan sulfate glycosaminoglycan (HS) chains. Dally localizes on the cell membrane through a glycosylphosphatidylinositol anchor and serves as a major Dpp/BMP co-receptor (Dejima et al., 2011;Nakato and Li, 2016). Dally forms a complex with Dpp and stabilizes it in the extracellular space, thus facilitating the formation of the Dpp morphogen gradient (Akiyama et al., 2008). A recent study demonstrates that Pent internalizes Dally and Dally-like, another Drosophila glypican, from the cell surface via a Rab5-dependent endocytosis, and this Pent-mediated degradation of glypicans plays a critical role in gradient formation (Norman et al., 2016). In addition, another extracellular Dpp/BMP feedback regulator, Ltl, physically interacts with Dally-like and antagonizes Dpp/BMP signaling in both the wing disc and the pupal wing (Szuperak et al., 2011).
Glypican activity is thus a key factor in orchestrating Dpp morphogen actions in the developing wing. In this study, we use a novel conditional mutagenesis approach to elucidate targets of Dpp signaling in the wing disc and identify the secreted Dpp/BMP feedback regulator Nord, a Drosophila Neuron-Derived Neurotrophic Factor (NDNF) homologue. nord mutants develop smaller wings and exhibit wing venation defects at low penetrance. Further consistent with a role in feedback regulation, overexpression of Nord antagonizes Dpp/BMP signaling. In addition, we show that Nord binds to Dally and that loss of nord suppresses the wing vein defects normally observed in dally mutants. Based on these data, we propose a model in which Nord functions with Dally to modulate Dpp/BMP signaling during wing development.

Dpp/BMP target genes in the developing wing disc
Due to a lack of techniques that completely disrupt dpp function, transcriptional targets of Dpp/BMP in the developing wing have been identified by relatively indirect approaches, such as genome-wide in silico screening of Dpp/BMP silencer elements and microarray comparison between wild-type cells and BMP type I receptor tkv mutant clones (Vuilleumier et al., 2010;Szuperak et al., 2011;Organista et al., 2015).
Recently, advances in genome editing technology have led to the creation of three dpp conditional knockout lines, dpp FO , dpp CA and dpp PSB (Akiyama and Gibson, 2015a;Bosch et al., 2017). Using flippase recombination target sequence (FRT)-mediated excision of the endogenous dpp locus, these lines allow for conditional elimination of dpp function in the wing disc and thus an opportunity to elucidate Dpp/BMP target genes using mRNA sequencing analysis.
Hypothetically, animals homozygous for conditionally inactivatable alleles should feature fully wild-type activity before FRT-mediated excision and null activity after. In order to identify the most suitable dpp conditional allele for RNA-Seq profiling, we first examined homozygous animals for dpp phenotypes and tested each line using conventional complementation testing ( Figure 1A and B). In the absence of conditional inactivation, we found that a majority of dpp PSB homozygotes (63%) displayed a posterior crossvein (PCV) phenotype, which was observed in just 3.8% of dpp CA adults and was absent in dpp FO flies ( Figure 1A). Since PCV development is sensitive to Dpp/BMP activity (Conley et al., 2000;Ralston and Blair, 2005;Serpe et al., 2005;Shimmi et al., 2005), this suggests that the genome editing of dpp PSB may mildly interfere with dpp function. dpp is known as a haploinsufficient gene, and its gene dosage is critical for viability (Irish and Gelbart, 1987;Wharton et al., 1996). Thus, to directly assess dpp activity in all three conditional knockout lines, we also performed an adult viability assay by crossing these conditional lines to weak (dpp hr56 ) and strong (dpp hr27 ) loss of function alleles ( Figure 1B) (Spencer et al., 1982;Irish and Gelbart, 1987;Wharton et al., 1996). Consistent with loss of dpp function in the absence of FRTmediated excision, dpp PSB /dpp hr56 and dpp PSB /dpp hr27 flies showed reduced adult viability (26.3% and 13.6% survival, respectively; Figure 1B). Unexpectedly, dpp CA , whose homozygotes rarely exhibited PCV defects, had the worst survival outcomes.
Only 4% of dpp CA /dpp hr56 animals survived to adulthood, and dpp CA completely failed to complement dpp hr27 ( Figure 1B). These results indicate significant loss of function prior to conditional inactivation. In contrast, dpp FO had the least impact on dpp activity. 40.1% of dpp FO /dpp hr27 animals survived to adulthood and dpp FO /dpp hr56 showed a similar viability to +/dpp hr56 controls (85.1%; Figure 1B). We therefore used the dpp FO line for further analysis.
As reported in previous studies, the endogenous stripe of Dpp expression is consistently eliminated in dpp FO /dpp FO ; dpp-Gal4/UAS-FLP (Dpp stripeless) animals via flippase-mediated recombination. This, in turn, results in loss of the Dpp/BMP activity (p-Mad) gradient in third-instar wing discs ( Figure 1C-F) (Akiyama and Gibson, 2015a).
We used this approach to discover Dpp/BMP target genes by comparing the transcriptomes of wild-type and Dpp stripeless wing discs, identifying 164 genes upregulated in the absence of dpp and 101 that were downregulated (an adjusted Pvalue < 0.01 and a fold change > 1.5; Figure 1G; supplemental excel file). Gene ontology (GO) term enrichment analysis revealed that genes regulating cell fate specification were enriched in the downregulated gene cluster, whereas upregulated genes were involved in biological processes such as transmembrane transport and cellular protein localization ( Figure 1H). Further, consistent with the mild growth phenotype of Dpp stripeless wing discs, cell cycle associated genes were not significantly enriched in either population ( Figure 1H). Validating our experimental approach, transcriptional profiling identified multiple well-known Dpp/BMP target genes, including the feedback regulators Dad, pent and ltl, ( Figure 1I) (Nellen et al., 1996;Tsuneizumi et al., 1997;Campbell and Tomlinson, 1999;Minami et al., 1999;Vuilleumier et al., 2010;Szuperak et al., 2011). In addition to known Dpp/BMP targets, we identified a novel Dpp/BMP target gene, nord ( Figure 1I).

Dpp/BMP regulates nord expression in the wing disc
We performed RNA fluorescent in situ hybridization (FISH) to visualize nord expression in wing discs and observed robust expression throughout the third-instar larval stage  (Nellen et al., 1996;Tsuneizumi et al., 1997;Campbell and Tomlinson, 1999;Minami et al., 1999;Vuilleumier et al., 2010;Szuperak et al., 2011). Further, in the pupal wing 24-  (Ito et al., 1997). As expected, we found that ectopic inhibition of Wg expression is high, we also examined whether ectopic Wg/Wnt pathway activation is sufficient to repress nord. We generated arm S10 FLP-OUT clones throughout the wing disc to activate Wg/Wnt signaling but found that nord expression was not affected (Figure 2-figure supplement 2G and H). Collectively, these results suggest that nord expression is not downstream of Wg/Wnt signaling and that Dpp/BMP signaling is required but not sufficient to drive nord expression in the wing disc.

nord governs growth and patterning during wing development
To determine if nord gene function is necessary for wing development, we used CRISPR/Cas9 mutagenesis to generate two deletion alleles, nord D1728 and nord D1 ( Figure   3A; Figure 3-figure supplement 1). nord encodes a secreted protein with an NDNF motif, which is essential for its function ( Figure 3A). Both deletions caused frameshifts, resulting in predicted protein truncations lacking large portions of the NDNF domain ( Figure 3A; Figure 3-figure supplement 1). Since the nord D1728 deletion resulted in an earlier termination codon compared to the nord D1 deletion, we mainly used this mutant allele for subsequent functional analyses. To determine how loss of nord affects Dpp/BMP signaling during wing development, we stained nord D1728 homozygous mutant discs with anti-p-Mad antibodies ( Figure 3B-D). We found that both the control and nord D1728 discs show similar p-Mad gradients, though the mutant discs had a slightly higher anterior p-Mad peak ( Figure 3B-D). Second, we checked Dpp/BMP activity in the 24-hour APF pupal wing. At this stage, p-Mad signal was detected in both future longitudinal veins and the crossvein regions ( Figure 3E and G). Although the majority of mutant pupal wings showed comparable Dpp/BMP activity to controls, they did exhibit ectopic Dpp/BMP activity in the PCV region at low frequency (13%; Figure 3F and H).
Finally, we investigated adult wing phenotypes ( Figure 3I-L). As might be expected based on the presence of ectopic Mad phosphorylation, nord D1728 homozygotes infrequently displayed a spur of ectopic venation in the PCV region (8.6%; Figure 3I-K).
Importantly, we also found that nord D1728 animals developed significantly smaller wings than the control ( Figure 3I, J and L). Complementation tests with a genomic deficiency confirmed these findings: nord D1728 /Df(2R)BSC356 and nord D1 homozygous animals show a similar PCV phenotype to nord D1728 homozygotes ( Figure 3K). These results rule out the possibility of off-target effects caused by CRISPR/Cas9-mediated mutagenesis and suggest that both nord mutants are likely null alleles.
BMP feedback regulation plays a critical role in buffering against both genetic and environmental perturbations (O'Connor et al., 2006;Lander, 2011;Raftery and Umulis, 2012;Bier and De Robertis, 2015). Therefore, we tested if nord could function to buffer Dpp/BMP signaling after genetic perturbations or when animals were kept under environmentally stressful conditions. Indeed, the PCV phenotype caused by a loss of nord was enhanced in both tkv (a Drosophila BMP type I receptor) and Mad (a Drosophila Smad1/5/8) heterozygous mutant backgrounds ( Figure 3K). Additionally, when nord mutant animals were reared at high temperature (29°C), the penetrance of the PCV defect was significantly increased ( Figure 3K). Based on these results combined with Nord's protein structure and its highly localized expression pattern during wing development, we propose that Nord acts as a secreted Dpp/BMP feedback regulator which functions to ensure robust developmental outcomes under sensitized genetic and stress conditions.

nord antagonizes Dpp/BMP signaling
To gain a better understanding of how Nord functions in Dpp/BMP signaling, we overexpressed nord using the Gal4/UAS system ( Figure 4) (Brand and Perrimon, 1993).
nord overexpression in the wing pouch domain using nub-Gal4 resulted in a reduced p-Mad gradient with a weaker posterior peak compared to controls ( Figure 4A-C). nord overexpression caused a variety of wing defects, including reduced size and a consistent loss of the anterior cross vein (ACV; Figure 4D-G). We also observed a PCV defect and wing notching at a lower penetrance ( Figure 4D-F). Although the crossvein defects are most likely caused by aberrant Dpp/BMP signaling activity, wing notching is linked to other pathways, including the Notch and Wg/Wnt signaling pathways (Zacharioudaki and Bray, 2014;Bejsovec, 2018). Therefore, we examined whether nord overexpression also affects Wg/Wnt signaling ( 1A-C). Altogether, our results demonstrate that nord overexpression not only antagonizes Dpp/BMP signaling but also inhibits the Wg/Wnt signaling pathway.

nord acts with HSPGs to modulate Dpp/BMP signaling
How does Nord regulate these signaling pathways in the wing disc? As mentioned above, nord encodes a secreted NDNF family protein predicted to possess heparinbinding activity in vertebrates (e.g. human; https://www.uniprot.org/uniprot/Q8TB73).
Heparin is a variant of HS polysaccharides, which are ubiquitously present in all tissue types as a major component of HSPGs (Hammond et al., 2014;Nakato and Li, 2016;Xie and Li, 2019;Jayatilleke and Hulett, 2020). In the developing wing disc, Dally and Dally-like, members of the glypican family of HSPGs, regulate both the Dpp/BMP and Wg/Wnt signaling pathways by acting as co-receptors (Nakato and Li, 2016;Mii and Takada, 2020). Thus, we hypothesized that Nord acts with HSPGs to regulate these signaling pathways. To test this, we first examined subcellular localization and heparin binding activity of Nord using Drosophila S2 tissue culture cells ( Figure 5A and B). S2 cells were transfected with nord-HA DNA to transiently express Nord-HA protein. We found that the signal peptide of Nord-HA protein was efficiently cleaved, and the mature Nord-HA protein was secreted into the conditioned media ( Figure 5A). As expected, the secreted Nord-HA protein also bound to heparin ( Figure 5A and B). Next, to test a physical interaction between Nord and Dally, a major co-receptor for Dpp/BMP signaling (Dejima et al., 2011), we performed co-immunoprecipitation experiments using S2 cell lysates containing Nord-HA with or without Dally-Myc. Dally-Myc protein was only recovered when S2 cells were co-transfected with both nord-HA and dally-Myc DNAs ( Figure 5C). Further, we found that this physical interaction was not totally dependent on HS because Nord also binds to Dally DGAG , a mutant form of Dally lacking an HS modification ( Figure 5C) (Kirkpatrick et al., 2006). Nevertheless, it seems likely that the presence of HS is able to enhance the physical interaction between Nord and Dally ( Figure 5C).  (Fujise et al., 2003;Akiyama et al., 2008). Although nord D1728 ; dally gem double mutant wing discs had similar p-Mad gradients to those observed in dally gem homozygotes, we observed that the majority of double mutant discs were able to specify the L5 vein primordia as visualized with anti-Delta (Dl) staining ( Figure 5D-J). As a result, the double mutants developed an intact L5 vein in both pupal and adult wings ( Figure 5K-Q). We also noticed that heterozygosity for nord D1728 was sufficient to partially rescue the L5 defect, suggesting a strong genetic interaction ( Figure 5Q). Further, the nord PCV defect was completely suppressed in nord D1728 ; dally gem double mutants, although wing notching associated with loss of dally was enhanced and the wing size defects were not rescued in the double mutant animals (Figure 5-figure supplement 1A-C). We therefore infer that loss of nord function both suppresses or enhances dally mutant phenotypes in a highly context-dependent manner. Altogether, these data suggest that Nord functions with the co-receptor Dally to modulate both Dpp/BMP and Wg/Wnt signaling.

The Dpp/BMP morphogen gradient and wing disc growth
While the Dpp/BMP morphogen activity gradient is clearly essential for wing patterning, the mechanistic basis for its requirement in wing disc growth remains poorly understood (Akiyama and Gibson, 2015a;Barrio and Milan, 2017;Bosch et al., 2017;Matsuda and Affolter, 2017). Here, we found that three independently constructed dpp conditional knockout lines displayed distinct growth phenotypes when the Dpp stripe expression was disrupted (Akiyama and Gibson, 2015a;Bosch et al., 2017;Matsuda and Affolter, 2017). dpp FO caused mild growth defects, whereas two additional lines, dpp CA and dpp PSB , showed severe growth impairments (Bischof et al., 2007;Akiyama and Gibson, 2015a;Matsuda and Affolter, 2017). These phenotypic discrepancies may be explained by the possibility that the dpp locus is less efficiently excised in dpp FO flies than in the other two lines (Bosch et al., 2017). In this study we found that the different genetic engineering strategies used to create these conditional alleles likely disrupt the unexcised dpp locus to different degrees ( Figure 1A and B). In addition to the FRT insertion, the dpp CA and dpp PSB alleles carry a Hemagglutinin (HA) tagged dpp and a visible selection marker (3xP3-RFP) in close proximity to the dpp locus (Bosch et al., 2017). These insertions might create a weakly sensitized genetic background that explains the different wing growth phenotypes. It is also possible that HA-tagged Dpp is less stable or cannot activate the signaling pathway as efficiently as the wild-type Dpp.  (Akiyama et al., 2012). Therefore, we infer that Dpp protein perdurance and/or residual Dpp expressing cells may be enough to sustain tissue growth until the late third larval instar without the canonical Dpp/BMP activity gradient (Akiyama and Gibson, 2015a). Consistent with this, it has been reported that uniform Dpp expression is able to maintain tissue growth in the wing disc devoid of the stripe Dpp expression, indicating a Dpp/BMP activity gradient independent growth mechanism (Bosch et al., 2017).
We found that Dpp/BMP signaling activity is essential but not sufficient to induce nord suppression at the D/V boundary remains elusive. Interestingly, dally expression is also controlled by these signaling pathways (Fujise et al., 2003). dally expression is activated by Hh signaling in the central stripe region of the wing disc, while Dpp/BMP signaling downregulates dally expression outside the stripe domain. Thus, through transcriptional regulation, nord and dally are co-expressed in part of the central stripe region. This may allow Nord to efficiently modulate functions of HSPG Dally to fine tune the Dpp/BMP signaling pathway in the extracellular space.

Extracellular HSPG regulation by Nord
Extracellular regulation of HSPG, such as shedding of membrane-localized HSPGs from the cell surface, and digestion and/or modification of HS chains, is tightly regulated during development and homeostasis, and its misregulation has been recognized as a hallmark of cancer (Hammond et al., 2014;Nakato and Li, 2016;Xie and Li, 2019;Jayatilleke and Hulett, 2020). Matrix metallopeptidase 9 cleaves a transmembrane form HSPG syndecan-1 from the cell membrane, facilitating tumor growth, metastasis and angiogenesis (Yang et al., 2007;Purushothaman et al., 2008). In addition, the Drosophila endo-6-O-suflatase Sulf1, which selectively removes 6-O-sulfate groups from the HS chains in the extracellular space, controls molecular interactions between HSPGs and signaling molecules including BMP and Wg, and regulates diverse biological processes such as wing development and intestinal stem cell activity (Kleinschmit et al., 2010;Wojcinski et al., 2011;You et al., 2011;Butchar et al., 2012;Dani et al., 2012;Kleinschmit et al., 2013;Takemura and Nakato, 2017). It has also been reported that a secreted Dpp/BMP feedback regulator, Pent, degrades glypicans by endocytosis for the proper formation of Dpp and Wg morphogen gradients (Norman et al., 2016). Here, we showed that Nord regulates Dpp/BMP signaling as an antagonist Since Dpp is also a heparin-binding protein and interacts with Dally, we favor a model in which Nord controls Dpp/BMP signaling by modulating co-receptor Dally availability on the cell surface ( Figure 5R) (Akiyama et al., 2008). In this framework, Nord competes with Dpp for Dally binding, thus antagonizing Dpp/BMP signaling. Further, a recent study demonstrates a physical interaction between Nord and Dpp (X. Zheng and MB. O'Connor, personal communication and coordinated submission). Therefore, Nord could also regulate Dpp/BMP signaling by directly interacting with Dpp ( Figure 5R).

Implications and conclusions
nord encodes a Drosophila NDNF protein. Human NDNF acts as a tumor suppressor (Xia et al., 2019;Zhang et al., 2019). NDNF expression is decreased in lung adenocarcinoma, and downregulation of NDNF promotes tumor growth in a mouse xenograft model (Zhang et al., 2019). NDNF is also known as a causative gene for congenital hypogonadotropic hypogonadism (CHH), which is characterized by infertility and delayed/absence of puberty (Messina et al., 2020). CHH is rooted in abnormal development of gonadotropin-releasing hormone (GnRH) neurons. Many CHH-linked genes are involved in the regulation of the Fibroblast growth factor (FGF) signaling pathway (Neocleous et al., 2020). Indeed, NDNF overexpression inhibits FGF signaling in cell culture, and ndnf mutant mice exhibit a GnRH neuronal migration defect (Messina et al., 2020). Further, NDNF improves cardiac function after ischemia and myocardial infarction by reducing cardiomyocyte apoptosis and promoting angiogenesis through activation of the AKT signaling pathway (Ohashi et al., 2014;Joki et al., 2015;Song et al., 2017). Nevertheless, the molecular mechanisms by which NDNF modulates multiple signaling pathways remain unclear. Intriguingly, it has been shown that HSPGs play essential roles in these biological processes (Purushothaman et al., 2008;Hammond et al., 2014;Jayatilleke and Hulett, 2020). For instance, mutations in heparan sulfate 6-O-sulfotransferase-1 and anosimin-1 are found in CHH patients (Tornberg et al., 2011;Hu et al., 2013;Neocleous et al., 2020). Anosmin-1 is an HSPG interacting protein and regulates GnRH neuronal migration by promoting FGF signaling (Hu et al., 2013). Taken together, we speculate that vertebrate NDNF may modulate HSPG activity to fine tune multiple downstream signaling pathways. In this regard, the present study not only reveals a new mode of extracellular morphogen regulation, but also opens up a new avenue for investigating the molecular basis for NDNF-associated disorders in humans.

Identification of Dpp/BMP target genes and gene ontology analysis
Wing discs from Oregon-R (wild-type) and dpp FO /dpp FO ; dpp-Gal4/UAS-FLP (Dpp stripeless) larvae were dissected. Total RNA from wing discs was prepared using Direct-zol RNA Miniprep Plus (R2071, Zymo Research). After confirming RNA quality by Bioanalyzer RNA Analysis (Agilent), sequencing libraries were generated using TruSeq RNA Library Prep Kit (Illumina) and sequenced on an Illumina HiSeq system (single read, HiSeq-50bp). 265 genes were differentially expressed in Dpp stripeless wing discs compared to wild-type samples (an adjusted P-value < 0.01 and a fold change > 1.5).
Gene ontology term enrichment analysis of 265 genes was performed on the PANTHER Classification System web site (http://pantherdb.org) (Mi et al., 2019) nord RNA FISH 1,029-bp nord cDNA fragment with T3 promoter was amplified from RE56892 clone (DGRC#9626) using primers 1 and 2 (Supplementary Table) and used for making a Digoxigenin (DIG)-labeled FISH probe using T3 RNA polymerase (P2083, Promega).
After T3 RNA polymerase reaction at 37°C for 4 hours, the reaction mixture was treated with RQ1 RNase free DNase (P6101, Promega) at 37°C for 30 minutes to digest template DNAs.
Female third-instar wing discs and 24-hour APF pupal wings were dissected in cold Schneider's media and fixed with 4% paraformaldehyde (PFA)/phosphate buffered saline (PBS). Wing discs were fixed at room temperature (RT) for 20 minutes and a pupal wing fixation was conducted at 4°C for three days. Blocking/PBST at RT for 2 hours and incubated with anti-DIG-POD (1:1,000, 11207733910, Roche) at 4°C overnight. After PBST and TNT washes at RT for 20 minutes three times each, nord mRNAs were visualized using TSA Plus Cyanine 5 (NEL745001KT, PerkinElmer). Dpp and GFP expression after nord FISH were detected as described below in the Immunohistochemistry and imaging section.

Generation of nord mutants
To generate nord D1 and nord D1728 , we created two sgRNA DNA constructs using primers 3-6 listed in Supplementary Table. Primers were annealed and cloned into the BbsI site of pBFv-U6.2 (Kondo and Ueda, 2013). A DNA mixture containing two nord sgRNA and a white sgRNA DNA constructs (100 ng/µl for each) was injected into the posterior region of embryos expressing Cas9 under the vas promoter (BDSC#66554). A white eye phenotype was used as a Cas9 activity indicator. Mutant candidates were screened by MiSeq analysis, and the lesions of two mutant lines were reconfirmed by Sanger sequencing.

Drosophila crosses
Adult viability experiment of dpp conditional knockout lines ( Figure 1B) (1:500 in PBST, Thermo Fisher Scientific) at RT for 3 hours. Samples were washed with PBST as described above and mounted using ProLong Gold Antifade Mountant (P10144, Thermo Fisher Scientific). All images were obtained using a Leica TCS SP5 confocal microscope.
Adult wings were dehydrated using pure ethanol and mounted with Leica Micromount (3801731, Leica). All pictures were taken by a Leica CTR 5000.

p-Mad and fz3-RFP intensity plot profiles and wing size
All wing disc images were collected at the same confocal setting on the same day and analyzed by the RGB profiler of FIJI. The "Measure" function of FIJI was used to analyze sizes of wing discs.

Cellular fractionation
pAWH-nord was generated by the GATEWAY system (Thermo Fisher Scientific). nord cDNA amplified from RE56892 clone (DGRC#9626) using primers 9 and 10 was cloned into pDONR221, a GATEWAY Entry vector, via the BP reaction (11789020, Thermo Fisher Scientific). After confirming nord cDNA sequence in the Entry vector, the cDNA fragment was subcloned from pDONR221 to pAWH (a GATEWAY Destination vector for C-terminal 3xHA tagging, DGRC#1096) using LR Clonase II Enzyme mix (11791020, Thermo Fisher Scientific), resulting in pAWH-nord.
Drosophila S2 cells were transfected with pAWH-nord. After 72-hour incubation at 25°C, the cell culture was collected and used as a total fraction. To prepare cell and medium fractions, the culture was centrifuged at 2,000 rpm for 10 minutes at 4°C.

Heparin-binding assay
Conditioned media from S2 cells transfected with pAWH-nord were applied to a 1 mL Hitrap Heparin HP column (17040701, Cytiva). The column was washed with 10 bed volumes of PBS. Heparin-binding proteins were eluted with 1M NaCl/PBS.

S2 cell-based BMP signaling assay
S2 cell-based signaling assay utilized a lacZ reporter whose expression was induced by Notch activation via S(H) binding sites and repressed by BMP silencer elements (Muller et al., 2003;Akiyama et al., 2012). S2 cells were transfected with 10 ng of pAW-dpp or pAW-dpp 3xHA along with the lacZ reporter and other DNAs required for this assay.
After 72 hour-incubation at 25°C, S2 cells were lysed to measure b-galactosidase activity using the Dual-Light Luciferase & b-Galactosidase Reporter Gene Assay System (T1003; Thermo Fisher Scientific). Luciferase activity was used for normalizing transfection efficiency.        (A-H) nord RNA FISH in the wing discs carrying control, tkv QD , Dad and arm S10 overexpressing clones. While Dad overexpression strongly reduced nord expression (C, D; arrow), nord expression was not clearly affected in tkv QD and arm S10 clones (E-H).
Scale bar: 50 µm for A. Anterior is oriented to the left.     Signaling activity of wild-type Dpp and Dpp 3xHA were examined using an S2 cellbased BMP signaling assay. Dpp 3xHA showed a reduced signaling activity compared to Dpp.