A drug repurposing screen reveals dopamine signaling as a critical pathway underlying potential therapeutics for the rare disease DPAGT1-CDG

Abstract

DPAGT1-CDG is a Congenital Disorder of Glycosylation (CDG) that lacks effective therapies. It is caused by mutations in the gene DPAGT1 which encodes the first enzyme in N-linked glycosylation. We used a Drosophila rough eye model of DPAGT1-CDG with an improperly developed, small eye phenotype. We performed a drug repurposing screen on this model using 1,520 small molecules that are 98% FDA/EMA-approved to find drugs that improved its eye. We identified 42 candidate drugs that improved the DPAGT1-CDG model. Notably from this screen, we found that pharmacological and genetic inhibition of the dopamine D2 receptor partially rescued the DPAGT1-CDG model. Loss of both dopamine synthesis and recycling partially rescued the model, suggesting that dopaminergic flux and subsequent binding to D2 receptors is detrimental under DPAGT1 deficiency. This links dopamine signaling to N-glycosylation and represents a new potential therapeutic target for treating DPAGT1-CDG. We also genetically validate other top drug categories including acetylcholine-related drugs, COX inhibitors, and an inhibitor of NKCC1. These drugs and subsequent analyses reveal novel biology in DPAGT1 mechanisms, and they may represent new therapeutic options for DPAGT1-CDG.

Author summary

Some proteins require the attachment of sugars—like mannose and galactose—in order to perform their function. The process that attaches these sugars is called "glycosylation". Genetic mutations in glycosylation pathways cause Congenital Disorders of Glycosylation ("CDGs"). CDGs are devastating disorders causing a multitude of symptoms, including seizures and developmental delay, and they have few treatment options. We wanted to find new potential treatment options for the CDG called DPAGT1-CDG. We use a model of DPAGT1-CDG that has inhibited glycosylation resulting in impaired eye development. We used a "drug repurposing screen", where we screened 1,520 drugs that are 99% FDA- or EMA-approved, to find drugs that improved its eye. We found 42 drug hits, including classes like dopamine receptor inhibitors, NSAIDs, and antihistamines. We found that the dopamine receptor inhibitors caused the strongest improvement. In addition, genetically impairing multiple parts of the dopamine pathway (e.g. how dopamine is made or recycled) also improved the model. This suggests that dopamine signaling is important in DPAGT1-CDG, and it could represent a new therapeutic option for patients. Overall, we present validation of several of our drugs, and these represent potential new treatments while also improving our understanding of glycosylation.

Introduction

Nearly 10% of the US population has a rare disease, yet over 95% lack effective therapeutics [1,2]. One solution to find new therapeutics is drug repurposing which use drugs that are already approved or under investigation [3]. When successful, drug repurposing screens provide immediate potential therapeutics that have already passed the rigor of human safety trials and may have a faster route of clinical approval [3]. Alternatively, they can potentially be used in an off-label capacity at a clinician’s discretion [3–5]. Even if a repurposed screen does not immediately result in a new therapy, it can still identify new biological interactions that can increase our understanding of a disorder.

DPAGT1-CDG is a rare disease that results from autosomal recessive loss-of-function mutations in the gene encoding the enzyme DPAGT1. It is one of nearly 200 Congenital Disorders of Glycosylation (CDGs) [6–9]. Glycosylation includes multiple pathways where sugars are co- or post-translationally added to proteins, lipids, or RNAs [10,11]. For proteins, it is critical for their proper localization, folding, and function [10]. DPAGT1 is part of the N-glycosylation pathway where these sugars are added to specific asparagine (N) residues, typically part of the canonical consensus sequence N-X-S/T where X is a non-proline amino acid [12]. DPAGT1 synthesizes dolichol-PP-GlcNAc which is the first step in N-glycosylation [6].

DPAGT1-CDG causes developmental delay, muscle weakness, and seizures, among other symptoms [6,13,14]. Less severe mutations in DPAGT1 can cause a form of congenital myasthenia syndrome (CMS) called DPAGT1-CMS [15]. Unlike DPAGT1-CDG, the CMS form has a known disease mechanism caused by hypoglycosylation of acetylcholine and calcium receptors [15,16]. Acetylcholinesterase inhibitors that increase acetylcholine levels can alleviate muscle weakness symptoms in both DPAGT1-CDG and -CMS patients [17–19]. For treating seizures, antiepileptic drugs have also been prescribed for DPAGT1-CDG and other CDGs [20]. However, both of these treatments are palliative, and there remains a great need for better therapeutics for the multisystemic symptoms in DPAGT1-CDG.

Drosophila share ~75% of human disease-causing genes [21] and have been regularly used in drug screens [22–24]. Drug repurposing screens in Drosophila have had previous success in identifying new potential therapies for CDGs. For example, our laboratory identified GSK3β inhibitors as a potential treatment for NGLY1 deficiency [25]. In addition, a drug repurposing screen in C. elegans identified the aldose reductase inhibitor epalrestat for treating PMM2-CDG [26,27]. Thus, a drug repurposing screen in Drosophila for DPAGT1-CDG could help find new potential therapeutics for this disorder.

In this study, we performed a drug repurposing screen on a Drosophila model of DPAGT1-CDG to identify new therapeutics and biological interactions with DPAGT1. We identified 42 compounds that suppress the DPAGT1-CDG model and genetically confirmed many of the drug interactions. These 42 compounds include acetylcholine-related drugs, dopaminergic antagonists, and cyclooxygenase inhibitors, among others. Importantly, we find that both pharmacologic and genetic inhibition of the D2 receptor can partially rescue our DPAGT1 model. In addition, impairing dopamine (DA) synthesis or recycling can also partially rescue the model. Our data suggest that manipulating DA signaling is a promising therapy for DPAGT1-CDG and implicates a larger role of DA signaling on N-glycosylation. The compounds and pathways identified in this screen are potential new therapeutic and genetic interactions with DPAGT1 that may represent new treatment options for DPAGT1-CDG.

Results

Identifying suppressors of a DPAGT1-CDG model in a primary drug repurposing screen

To model DPAGT1-CDG, we used an eye-based Drosophila model that we previously established [28]. The Drosophila eye is regularly used to study and model biological processes, including development [29–32]. In our model, DPAGT1 expression is reduced in the eye by RNA interference (RNAi, UAS-GAL4 system [33], eya composite-GAL4 driver [34]). Analysis of whole fly heads of the DPAGT1 model resulted in a ~65% knockdown of DPAGT1 expression (S1A Fig). The compound eye makes up a large portion of the Drosophila head, but this knockdown is likely even lower because our RNAi is eye-specific. In vitro analysis of DPAGT1-CDG patient proteins found they had 10–50% protein activity [35]. Thus, having at most ~35% of normal DPAGT1 expression is a reasonable amount of knockdown to model DPAGT1 deficiency. This knockdown results in a small, improperly developed, rough eye phenotype (hereafter referred to as “DPAGT1 model”) [28] (Fig 1). To validate this small eye phenotype, we also created a second model using a different RNAi line (BDSC 51869) driven by eya composite-GAL4. This model has a small, rough eye phenotype comparable to our original model (S1B Fig).

Fig 1. Summary of the drug repurposing screen.

(A) Screening method for the drug repurposing screen. (B) Z-score plot of all repurposed drugs tested. Each point is the Z-score of a drug compared to the average of DMSO-treated control flies. Dotted lines indicate our Z-score threshold of 1.5 or -1.5. We found 42 and 16 drugs with a Z-score of ≥1.5 or ≤-1.5, respectively. (C) Representative images showing male Drosophila stocks of the eya composite-GAL4 driver (control, without RNAi), the RNAi background control to the DPAGT1 model (attP2), the DPAGT1 RNAi alone (control, without a GAL4 driver), the DPAGT1 model, and the effects of the top suppressor (Z-score = 2.85) and top enhancer (Z-score = -2.92). See S1 and S2 Tables for a complete list of compounds and Z-scores, and S2 Fig for representative female images.

The DPAGT1 model can be chemically or genetically manipulated to make the eye larger (suppress the phenotype) or smaller (enhance the phenotype) [28]. To find potential therapeutics, we used the Prestwick Chemical Library, a collection of 1,520 compounds that are 98% FDA- or EMA-approved. We raised DPAGT1 model flies on food containing each drug and measured the eye size of resulting progeny (Fig 1A). We used a 5 μM drug concentration—a standard dosage for flies that does not cause toxicity [25]. We scored 8,076 flies in total in the screen (4.4 average flies per treatment), and we calculated the Z-score of each compound by comparing their eye sizes to DMSO-treated DPAGT1 model control flies. We identified 42 compounds that partially suppressed the eye phenotype of the DPAGT1 model ("suppressors") (Fig 1B and 1C, S1 and S2 Tables) and 16 compounds that enhanced the eye phenotype ("enhancers") (Fig 1B and 1C, S1 and S2 Tables). 16 compounds resulted in no flies (S1 Table). Our positive hit rate of 2.8% falls in line with other compound screens performed in Drosophila [23,25,36,37].

We focused our validation efforts on suppressors that had ≥3 drugs in a particular drug class, had the strongest suppressive effect on eye size (Z-score ≥2), or had high potential for patient use (e.g. available over-the-counter). Drug classes with the highest representation were acetylcholine-related drugs, dopamine (DA) receptor antagonists, COX inhibitors, and antioxidants (three drugs each, S2 Table). In total, we tested 20/42 (48%) of the suppressors in validation experiments.

Impairing acetylcholine breakdown improves the DPAGT1 model

Mutations in DPAGT1 can cause muscle weakness because acetylcholine receptors (AChRs) become hypoglycosylated [15,16]. This muscle weakness can be treated by cholinesterase inhibitors which increase available acetylcholine [15,16,38]. Our unbiased screen identified three AChR-related drugs. This included methacholine, an analog of acetylcholine and a muscarinic receptor agonist [39], as well as two drugs expected to increase acetylcholine levels: benactyzine, a cholinesterase inhibitor/anticholinergic [40], and neostigmine, a cholinesterase inhibitor [41]. To validate drug hits from the screen, we primarily used drugs purchased from a second source to reduce potential quality control errors (this is true of all further validation) (S2 Table). In dose-response analyses, benactyzine and neostigmine both increased the eye size in the DPAGT1 model (Fig 2A and 2B). Benactyzine significantly increased female eye size at 25 μM and generally increased the upper distribution of eye sizes at most doses tested in both sexes (Fig 2A). Neostigmine significantly increased eye size in males at 5 and 25 μM (Fig 2B) and had a trending (+12%), but not statistically significant, increase in median size in females at 25 μM. As acetylcholinesterase inhibitors are already used in patients, these two drugs represent a strong validation of the screen to find patient-relevant hits.

Fig 2. Pharmacological and genetic inhibition of acetylcholinesterase improved the DPAGT1 model.

(A-D) Multiple ACh-related drugs can partially rescue the DPAGT1 model. For Neostigmine, we assayed 1 μM separately, but it was not significant. (E) RNAi knockdown of ACE also partially rescues the DPAGT1 model (VDRC 105432KK). See S2 and S3 Tables for more details. * p<0.05, ** p<0.01, *** p<0.001 (Student’s t-test or One-way ANOVA with multiple comparison correction).

We next assayed four acetylcholine-related drugs already used by DPAGT1-CDG or -CMS patients for their effect on the DPAGT1 model. This included pyridostigmine (used in DPAGT1-CMS [17,19] and DPAGT1-CDG patients [18]), amifampridine, salbutamol, and edrophonium (all used in DPAGT1-CMS patients, though edrophonium is more diagnostic [17,42]). Treatment with edrophonium strongly improved female DPAGT1 model flies in a dose-dependent manner (Fig 2C). Similar to its analog neostigmine (Fig 2B), treatment with pyridostigmine caused a significant increase in eye size in males at 5 μM (Fig 2D) and a small, but not significant increase in females at 25 μM (+6%). Overall, acetylcholine-affecting drugs identified in the screen and currently used by patients improved the DPAGT1 model.

The rescuing drugs neostigmine, pyridostigmine, and edrophonium all inhibit the enzyme acetylcholinesterase (AChE) which breaks down acetylcholine [16]. We tested whether genetic inhibition of the AChE-encoding gene ACE (human: ACHE) could partially rescue the model as well. We used two RNAi lines against ACE (via the GAL4/UAS system [33]) to mimic pharmacological inhibition. One previously studied [43] RNAi line improved the DPAGT1 model in both sexes (Fig 2E), mimicking the drug inhibition of ACE. The second RNAi line was capable of improving females (S3 Table). ACE knockdown on its own caused no effect in males, and only a 4–5% increase in females (see all eya composite-GAL4 control data in S3 Table). These pharmacological and genetic data indicate that inhibition of AChE, and the ostensible increase in acetylcholine, can partially rescue the eye development defect of the DPAGT1 model. This validates the screen, supports continued usage of AChE inhibitors in patients, and might also indicate a role for AChE inhibitors outside of their effect on muscle weakness.

Antagonizing D2 receptor signaling significantly improves the DPAGT1 model

DA synthesis, transport, and downstream receptor signaling is well-conserved between Drosophila and humans [44,45] (Fig 3A). In this section, we primarily discuss female data, as females had an overall more robust response to manipulating DA signaling. All results, including male data, can be found in S2 and S3 Tables. Three DA receptor antagonists were suppressors with moderate Z-scores in our screen (Z-scores: 1.8–1.86, S1 Table). These included sulpiride, a D2/D3 receptor antagonist [46], paliperidone, an antipsychotic that antagonizes D2 and 5-HT2A receptors [47], and prochlorperazine, a first-generation antipsychotic that primarily antagonizes D2 receptors but also targets adrenergic, cholinergic, and histaminergic receptors [48–50]. Of these, validation testing of prochlorperazine improved the DPAGT1 model at 1 μM in females (Fig 3B), and paliperidone also improved male flies (S2 Table). While evaluating our drug classes, we noticed that one D2 receptor antagonist—trifluoperazine [51]—had no observed progeny in the original screen (S1 Table). This is sometimes due to drug lethality like the insecticide fipronil (S1 Table), but it can also cause a false negative if parental flies did not mate well. Retesting trifluoperazine resulted in progeny and a robust increase in eye size in the DPAGT1 model at the 5 μM dose in both sexes (Fig 3C and S2 Table). While prochlorperazine, paliperidone, and trifluoperazine overlap in function on D2 receptors, they can also target other receptors and pathways. Given this, we focused on DA signaling by genetically reducing the expression of their shared target Dop2R (human: D2R) in the DPAGT1 model.

Fig 3. Inhibiting the dopamine 2 receptor improved the DPAGT1 model.

(A) A schematic of DA synthesis, recycling, metabolism, and signaling in Drosophila. The effect of each RNAi or mutation on the DPAGT1 model is indicated. Genes in blue with the "+" symbol improved the DPAGT1 model when knocked down, while those in red with the "-" symbol made it worse. Genes in black with the "0" symbol had no effect or could not be distinguished from the knockdown effect on its own. (B-C) The D2 antagonists prochlorperazine and trifluoperazine improved eye size in the DPAGT1 model. For technical reasons, we assayed 25 μM separately, but neither drug was significant. (D) RNAi knockdown of the Dop2R gene (BDSC 36824) improved the DPAGT1 model. (E) Representative images comparing the DPAGT1 model crossed with the control RNAi background (attP2, BDSC 36303) or a Dop2R RNAi line (BDSC 36824). (F) DPAGT1 model flies carrying a heterozygous null of Dop2R had increased eye size (BDSC 84720). (G) DPAGT1 model flies crossed to a Dop2R overexpression line (UAS-Dop2R, BDSC 86134) had worse eye size. (H) A heterozygous null of Dop1R1 (BDSC 92640) results in worse eyes in the DPAGT1 model. (I) Dop2R RNAi (BDSC 36824) improved DPAGT1 model [Dop1R1^+/-] flies, but to a smaller maximal eye size than in DPAGT1 model flies alone (indicated by dashed line, averaged from Fig 3D). All graphs are of female flies. See S2 and S3 Tables for more details. * p<0.05, *** p<0.001, **** p<0.0001 (Student’s t-test or One-way ANOVA with multiple comparison correction).

Dop2R is the only Drosophila ortholog of human genes D2R, D3R, and D4R. These encode for "D2-like" G-protein coupled receptors (GPCRs) that inhibit the adenylate cyclase/cAMP signaling pathway [44,45] (Fig 3A). Matching the pharmacological data from the primary screen, reduced expression of Dop2R using RNAi resulted in a strong improvement of eye size across all RNAi lines and replicates tested (Fig 3D and 3E and S3 Table), compared to other tested drug and RNAi treatments. In addition, their eyes showed strong qualitative improvement as they had fewer "glassy" eye sections (Fig 3E). We also crossed a heterozygous Dop2R null allele [52] into the DPAGT1 model (note that Dop2R is X-linked and only females are analyzed here). Mimicking the RNAi, DPAGT1 model flies carrying a heterozygous null Dop2R mutation also resulted in improved eye size (Fig 3F). This indicates that this improvement is not due to any indirect UAS-GAL4 [33] effects. Finally, overexpressing Dop2R [52] in the DPAGT1 model decreased eye size (Fig 3G), confirming what is expected given the knockdown and null result. Genetically manipulating Dop2R had little effect on its own, with only two cases changing eye size: an increase in Dop2R OE females, and a minor decrease in Dop2R hemizygous null males (S3 Table). Overall, this indicates that reduction of D2 receptor signaling improves the DPAGT1 model.

After release into the synaptic cleft, DA can bind to several receptors in the fly. This includes the Dop2R receptor as well as the Dop1R1, Dop1R2, and DopEcR receptors [44,45]. Dop1R1 and Dop1R2 are orthologs of the human D1R and D5R genes. These encode "D1-like" GPCRs that activate the adenylate cyclase/cAMP signaling pathway [44,45] (opposite of "D2-like" receptors). Given the opposite downstream effect of these receptors compared to Dop2R, we hypothesized that they would have an opposite effect on eye size. We crossed heterozygous null mutations of Dop1R1 [53] and Dop1R2 [52] into the DPAGT1 model. Supporting our hypothesis, DPAGT1 model flies carrying a heterozygous null Dop1R1 mutation had worse eyes than the DPAGT1 model alone (Fig 3H and S3 Table). DPAGT1 model flies carrying a heterozygous null Dop1R2 mutation had only a mild effect in one female replicate and no effect in males (S3 Table).

Given the strong partial rescue effect of Dop2R RNAi (Fig 3D–3E), we tested whether this loss of Dop2R could partially rescue the loss of Dop1R1 in the DPAGT1 model. We created a recombinant line of the DPAGT1 model with the Dop1R1 null mutation (DPAGT1 model [Dop1R1^+/-]). Loss of Dop2R does partially rescue the DPAGT1 model [Dop1R1^+/-] eye size, but to a lesser extent than loss of Dop2R on its own (Fig 3I, p<0.0001, Student’s t test, Dop2R RNAi compared across experiments). This result is in line with the opposing effects of Dop1R1 and Dop2R on downstream cAMP signaling (Fig 3A).

The fourth DA receptor in Drosophila, DopEcR, has no clear human ortholog. Similar to Dop1R1, DopEcR activates the downstream adenylate cyclase/cAMP signaling pathway through DA binding. However, DopEcR can also activate the Mitogen-Activated Protein Kinase (MAPK) pathway through binding of the insect steroid ecdysone [44,45]. RNAi knockdown of DopEcR also improved eye size in the DPAGT1 model (S3 Table). Taken together, these data suggest that loss of DA signaling through Dop2R and DopEcR is beneficial to the DPAGT1 model, while loss of DA signaling through Dop1R1 is detrimental to the model.

Impairing DA synthesis and recycling can improve the DPAGT1 model

DA is synthesized by the enzymes tyrosine hydroxylase and DOPA decarboxylase, which are encoded by the ple and Ddc genes, respectively (humans: TH and DDC). DA is then transported into vesicles by the vesicular monoamine transporter VMAT, encoded by Vmat (humans: VMAT2), before being shuttled into the synaptic cleft [44,45] (Fig 3A). Similar to knockdown of Dop2R, knockdown of the first DA synthesis gene, ple, and the second DA synthesis gene, Ddc, strongly improved the DPAGT1 model (Fig 4A and 4B and S3 Table). Knockdown of Vmat also improved the model, but it had a similar increase on its own (S3 Table). Overall, this suggests that inhibiting DA synthesis is beneficial to the DPAGT1 model.

Fig 4. Inhibiting DA synthesis and recycling improves the DPAGT1 model.

(A-B) RNAi targeting the genes encoding the DA synthesizing enzymes ple and ddc (BDSC 25796 and 27030) improved the DPAGT1 model. (C-D) RNAi against the gene encoding the DA recycling protein DAT (BDSC 50619), or pharmacologic inhibition of DAT via nomifensine at 1 μM, improved the DPAGT1 model. (E) RNAi against the gene encoding the DA recycling protein ebony (BDSC 28612) improved the DPAGT1 model. (F) RNAi against the gene encoding the N-acetyltransferase speck resulted in worse eyes in the DPAGT1 model. All data are of female flies. See S2 and S3 Tables for male data and more details. * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001 (Student’s t-test or One-way ANOVA with multiple comparison correction).

There are several ways that DA can exit the synaptic cleft. DA can be recycled back from the synaptic cleft to the presynaptic neuron by the DA transporter DAT [44,45] (encoded by DAT) (Fig 3A). Knockdown of DAT strongly improved DPAGT1 model eye size, matching the same increase as loss of Dop2R (Fig 4C, S3 Table). The DAT inhibitor [54,55] nomifensine maleate had a positive Z-score in the primary screen (0.84) but did not reach our hit threshold. Because we saw strong improvements in eye size when DAT was knocked down compared to other treatments, we tested a version of this drug lacking the maleate salt (S2 Table). In line with the genetic data, treatment with nomifensine improved the DPAGT1 model at 1 μM (Fig 4D). In Drosophila, but not in humans, DA can also be converted to the metabolic product NBAD for translocation and conversion back into DA. These steps are performed by the enzymes black, ebony, and tan through the β-alanylation pathway [44,45] (encoded by black, ebony, and tan) (Fig 3A). ebony and tan convert and shuttle DA, while black creates the β-alanine necessary for ebony to function [44]. RNAi knockdown of the first β-alanylation gene, ebony, strongly improved the DPAGT1 model (Fig 4E and S3 Table). However, while knockdown of black improved the DPAGT1 model, it had a similar increase on its own (S3 Table). There are no available RNAi lines for tan. Overall, inhibiting the recycling of DA back into the synaptic cleft through DAT or ebony improves the DPAGT1 model.

In contrast to these recycling pathways, DA can exit the synaptic cleft through metabolism into the inactive compound NADA in Drosophila. This occurs through the N-acetyltransferase enzyme, speck, and prevents further signaling by that DA molecule. While speck has no clear human ortholog, it serves a similar role to monoamine oxidases in humans [44]. Because speck removes DA, we hypothesized that speck would have the opposite effect of the DA synthesis genes. Supporting this hypothesis, knockdown of speck resulted in worse eye sizes in the DPAGT1 model (Fig 4F and S3 Table). Thus, inhibiting the removal of DA from the synaptic cleft is detrimental to the DPAGT1 model.

Generally, these results fit the hypothesis that most DPAGT1 model outcomes can be predicted by how a treatment affects DA binding to Dop2R (Fig 3A). For example, inhibiting DA synthesis or recycling may improve the DPAGT1 model by reducing the flux of DA to Dop2R. In contrast, because speck normally removes DA and reduces its binding to Dop2R, this could explain why inhibiting speck worsens the DPAGT1 model. Overall, these data strongly link the inhibition of DA synthesis, recycling, and signaling with improvement of DPAGT1 deficiency. Drugs that impair D2 signaling or synthesis, or are agonists of D1 receptors, may be good therapeutics for DPAGT1-CDG.

Histaminergic signaling is beneficial to the DPAGT1 model

The Histamine 2 (H2) receptor antagonist ranitidine, used to decrease stomach acid [56], was a strong enhancer in our screen (Z-score = -2.33, S1 Table). Histamine and DA have some overlapping biology. For example, Drosophila can process DA and histamine using the same β-alanylation enzymes—ebony, black, and tan [44,57,58]. In addition, loss of histamine can upregulate depolarization-stimulated DA release and receptor expression in mice [59]. Given the impact of DA signaling on the DPAGT1 model (Figs 3 and 4), our results from ebony knockdown (Fig 4E and S3 Table), and the finding of ranitidine, we tested histaminergic signaling further. In line with the primary screen, ranitidine significantly worsened eye size in the DPAGT1 model at multiple doses (Fig 5A). As ranitidine is an H2 receptor antagonist, we tested if exogenous histamine (an H1/H2 agonist) would have the opposite effect. Treatment with histamine showed statistically significant increase at 5 μM in males and increased the upper distribution of both sexes at 25 μM (Fig 5B). Histamine is an atypical medication which is used in select cancer therapies [60]. Histamine could represent a new therapeutic for patients, and these data also suggest that some antihistamines may be detrimental under DPAGT1 impairment.

Fig 5. Antagonizing histamine signaling worsens the DPAGT1 model.

(A) The H2 antagonist Ranitidine worsens the DPAGT1 model at multiple concentrations. (B) The H1/H2 agonist histamine can partially rescue the DPAGT1 model at 5 μM in males. See S2 Table for more details. * p<0.05, ** p<0.01 (One-way ANOVA with multiple comparison correction).

Inhibiting the ion transporter NKCC1 improves the DPAGT1 model

The loop diuretic bumetanide was a strong suppressor hit in our screen (Z = 2.04). Bumetanide inhibits the ion cotransporters NKCC1 and NKCC2 (encoded by NKCC1 and NKCC2 in humans). Both proteins transport Na+, K+, and Cl- ions into cells primarily in the secretory epithelia or renal tissues, respectively [61]. In addition to its diuretic property, bumetanide has recently been tested as an off-label anti-seizure medication to some success [62–65]. Finally, we previously found that NKCC1 is a genetic modifier of NGLY1 deficiency, another CDG [66].

Matching the primary screen, bumetanide caused a strong increase in eye size in the DPAGT1 model in a dose-dependent manner (Fig 6A). Ncc69 is the Drosophila ortholog of NKCC1 and NKCC2 [66,67]. Genetic knockdown of Ncc69 also resulted in an increase in eye size in the DPAGT1 model (Fig 6B and S3 Table). This corroborates with the bumetanide data that inhibiting NKCC1/NKCC2 activity is beneficial under DPAGT1 deficiency. Thus, drugs that inhibit NKCC1 activity may be beneficial when DPAGT1 activity is reduced.

Fig 6. Inhibiting the ion transporter NKCC1 improves the DPAGT1 model.

(A) Bumetanide increases eye size in the DPAGT1 model. (B) Genetic knockdown of Ncc69, which encodes the target of bumetanide inhibition, also increases eye size in the DPAGT1 model. See S2 and S3 Tables for more details. * p<0.05, **** p<0.0001 (Student’s t-test or One-way ANOVA with multiple comparison correction).

COX inhibitors improve the DPAGT1 model

The cyclooxygenase signaling pathway involves the synthesis of prostaglandins that are important for processes such as inflammation and growth [68–70]. Many non-steroidal, anti-inflammatory drugs (NSAIDs), such as ibuprofen, inhibit the first enzymes in this process—cyclooxygenases COX-1 and COX-2 (encoded by COX1 and COX2 in humans). There were three COX-1/COX-2 inhibitor hits in our screen: triflusal [71], antipyrine, and the antipyrine metabolic product, 4-hydroxyantipyrine [72]. Of these, antipyrine significantly increased average eye size at the 1 μM dose, and multiple antipyrine doses resulted in a positive shift in DPAGT1 model eye size distributions (Fig 7A).

Fig 7. Inhibiting COX enzymes improves the DPAGT1 model.

(A-B) Both antipyrine and tolmetin improved the eye size of the DPAGT1 model. (C) RNAi against the gene encoding the COX-like enzyme Pxt (BDSC 32382) partially rescued the DPAGT1 model. See S2 and S3 Tables for more details. * p<0.05, ** p<0.01, ** p<0.001, **** p<0.0001 (Student’s t-test or One-way ANOVA with multiple comparison correction).

NSAIDs are commonly used and would be easy to access for patients [73]. However, COX inhibitors can vary in their specificity and degree of COX inhibition, and this can affect their efficacy and tolerance [74–77]. As such, we tested five other COX inhibitors using the DPAGT1 model. Of these, the COX-1/COX-2 inhibitor tolmetin [78] significantly improved the DPAGT1 model at 5 μM in both sexes, and at 25 μM in females (Fig 7B). The Drosophila gene Pxt is the ortholog of human COX1 and COX2 [79,80]. Similar to antipyrine and tolmetin, knockdown of Pxt also resulted in improved eye size in the DPAGT1 model (Fig 7C and S3 Table). Given that inhibiting Pxt is beneficial, we hypothesize that antipyrine and tolmetin specifically improve the DPAGT1 model because they may be more efficacious or well-tolerated under DPAGT1 deficiency. Overall, inhibiting prostaglandin signaling through COX inhibition is beneficial to the DPAGT1 model, and certain COX inhibitors may be useful in alleviating symptoms of DPAGT1 impairment in patients.

Discussion

Here, we identify multiple drugs that improve a Drosophila model of DPAGT1 deficiency. Overall, 8/21 (38%) drugs tested from the primary screen validated in further dose-response analysis. In addition, 6/16 (38%) drugs derived from the primary screen improved the DPAGT1 model. Almost all tested drugs were genetically validated by manipulating their target genes in Drosophila. We found that drug classes involving acetylcholine signaling, dopamine (DA) synthesis, histamine receptors, NKCC1/2, and cyclooxygenase inhibition might be good candidates for treating DPAGT1-CDG (Fig 8).

Fig 8. New connections to DPAGT1 biology.

DPAGT1-CDG is caused by mutations in the enzyme DPAGT1. Here we summarize several new biological connections to DPAGT1 discussed throughout, along with potential specifics to each individual pathway in parentheticals.

One of the enriched classes was acetylcholine (ACh)-related drugs. Identifying this drug class provides strong validation of our fly model and drug screen, as this class is already used by both DPAGT1-CDG and -CMS patients [17,18,42]. In addition, multiple AChE inhibitors currently taken by patients improved the DPAGT1 model. This suggests that our CDG fly model is a reasonable representation of the disorder, and that these models are valuable tools for identifying potential therapeutics. We found that genetically or pharmacologically inhibiting acetylcholine breakdown is beneficial under DPAGT1 impairment. Acetylcholinesterase inhibitors are taken to improve muscle function in DPAGT1 patients by increasing acetylcholine signaling at neuromuscular junctions (NMJs) [81]. However, in Drosophila, glutamate is the excitatory neurotransmitter in NMJs [82], and the fly eye is likely not impacted by NMJ signaling regardless. Outside of NMJs, acetylcholine can also bind to nicotinic acetylcholine receptors to mediate DA release in flies [83] and mammals [84,85]. Given that increased DA appears negative for the model, it is possible that the benefit of increased acetylcholine signaling counteracts any negative consequences of increased DA release. Alternatively, acetylcholine-mediated DA release may be more prevalent in cells expressing higher levels of D1-like DA receptors, which would prove beneficial.

Acetylcholine is important for many biological processes, including neurodevelopment [86–88] and growth [89–91]. Given the impaired eye development in the DPAGT1 model, these processes may underlie the improvement from ACE inhibition (and the ostensible increase in acetylcholine). While muscle weakness in patients and the eye defect in the DPAGT1 model are distinct phenotypes, they are connected because both are improved by decreasing breakdown of acetylcholine. It is possible that impaired acetylcholine signaling may occur outside of muscles in DPAGT1-CDG patients. For example, patients also have impaired neurodevelopment as well as eye disorders such as cataracts [20,92]. Further use of AChE inhibitors may be worth exploring further in patients given the improvement in eye development in our model. Overall, acetylcholinesterase inhibitors have beneficial effects under DPAGT1 impairment in flies and humans.

We found that inhibiting Dop2R signaling improves the DPAGT1 model. In addition, knockdown of DA synthesis or recycling genes improves the model. Thus, the synthesis of DA, and its binding to the Dop2R, are detrimental under DPAGT1 inhibition. Dop2R signaling inhibits the downstream adenylate cyclase/cAMP signaling pathway. While cAMP signaling has many diverse functions, it is important for proper neuronal connectivity and survival [93,94]. Therefore, inhibiting Dop2R signaling may increase cAMP signaling and allow for improved neuronal outcomes. This could underlie the benefits in the DPAGT1 model from Dop2R genetic and pharmacological inhibition. Supporting this hypothesis, heterozygous knockout of the Dop1R1 receptor worsened the DPAGT1 model. Binding to Dop1R1 may counteract signaling to Dop2R through increased cAMP signaling. Given this, drug classes such as D2R antagonists or D1R agonists might be good therapeutics for DPAGT1-CDG. Of note, we identified two monoamine oxidase inhibitors (MAOIs) as enhancers in the screen (moclobemide and selegiline, S2 Table). MAOIs inhibit MAOs to prevent the breakdown of neurotransmitters such as DA [95]. It is possible that these MAOIs were enhancers because they increased the amount of DA in the presynaptic neuron, in line with our data on DA synthesis and recycling.

Misregulated DA signaling is associated with neurological disorders such as Parkinson’s disease, schizophrenia, and bipolar disorder [96]. DPAGT1-CDG has no previous clear connection to DA. However, both DPAGT1-CDG and -CMS patients have abnormal gait [42,97]. DPAGT1-CDG patients also have neurological symptoms including seizures and intellectual disability [6,14]. Misregulated DA could underlie some symptoms of DPAGT1-CDG. Interestingly, the DA pathway is misregulated in a fly model of another CDG, NGLY1 deficiency [25]. Further, aripiprazole, a complex agonist of D2 receptors [98], could rescue worm and fly disease models of NGLY1 deficiency [99]. Thus, DA signaling may be important in glycosylation disorders more generally.

Inhibiting the ion transporter Ncc69 via bumetanide or RNAi resulted in a strong improvement of the DPAGT1 model compared to other treatments. In humans, NKCC1 has three sites of N-glycosylation while NKCC2 has two (via Uniprot [100]). We previously found that loss of the deglycosylating enzyme NGLY1 results in reduced NKCC1 activity [66], and other studies find that inhibiting N-glycosylation reduces the function of NKCC1 and NKCC2 [101,102]. Therefore, proper glycosylation or deglycosylation of NKCC1 and NKCC2 is critical to their function. Given this, impaired N-glycosylation in the DPAGT1 model likely impairs the function of Ncc69. Thus, it was surprising that further reducing Ncc69 expression by RNAi actually improved the DPAGT1 model. We hypothesize that under DPAGT1 impairment, any Ncc69 that is expressed may be misglycosylated and subsequently may be misfolded, or have aberrant function, and contribute to disease pathogenesis.

Bumetanide is typically used as a loop diuretic in patients with edema and hypertension [65] due to its inhibition of NKCC2 in renal tissue. However, bumetanide also inhibits NKCC1 which is expressed more systemically. This includes neurons where NKCC1 affects cell polarization and levels of GABA that are important for proper CNS function [103–105]. To that end, bumetanide has been used more recently as an anti-seizure medication in mice [104] and off-label in humans [105]. In addition, oral solutions of bumetanide were recently used in clinical trials for autism in children [106]. However, bumetanide still impacts renal function when used for neurological symptoms, causing diuretic side effects [107]. These side effects are likely more challenging for those who are already medically fragile such as DPAGT1-CDG patients. However, the structural basis for bumetanide inhibition was recently described [65], and this could pave the way for future bumetanide-like drugs. It is possible that such drugs could be used as anti-seizure medications without the diuretic side effects. Given that bumetanide improved the eye development of the DPAGT1 model, it also suggests a role for bumetanide-like drugs beyond their use as anti-seizure medications.

The COX inhibitors tolmetin and antipyrine improved the DPAGT1 model. While now discontinued in the US (via PubChem [108]), tolmetin was used for decades as an alternative to aspirin and has similarly few side effects [109]. Antipyrine, also known as phenazone, is an antipyretic currently used to treat ear infections [110]. In the past, antipyrine was used orally for decades and was one of the earliest prescribed analgesics [111]. Interestingly, while the antipyrine metabolite, 4-hydroxyantipyrine, was a hit in the screen, it did not validate in our hands. This could indicate some volatility of this metabolite or that a different dose is required. Inhibiting the cyclooxygenases COX-1 and COX-2 is a common drug mechanism for alleviating inflammation and pain [68,69]. However, there is also evidence for the use of COX-1 and COX-2 inhibitors in treating cognitive deficits and convulsions [70,112,113], both of which are relevant to DPAGT1-CDG. While only two NSAIDs improved our model, this category may still be promising for DPAGT1-CDG given their high usage and ease of patient access [73].

We tested multiple drug doses to determine if some doses could provide better partial rescue of our model compared to our initial screen of 5 μM. In our model, some drugs may need high concentrations to elicit a phenotype (e.g. bumetanide), while others may have decreased efficacy at too high of a dose (e.g. nomifensine). This is a common finding in dose-response analyses [114]. Ultimately, as long as one dose worked, we considered it as a hit because many drugs require specific dose ranges to be effective [114]. We also found several drugs that partially rescue one sex stronger than the other (e.g. acetylcholinesterase inhibitors). Drosophila are sexually dimorphic and have differences in behavior, development, and immunity, among other pathways [115–117]. In addition, previously studied drugs such as rapamycin are only effective in one sex [118]. We chose to perform validation in both sexes to reduce potential sex-specific effects. In each drug category, at least one pharmacological or genetic treatment improved each sex. Thus, while there may be some sex differences in efficacy, we did not find any completely sex-dependent biological pathways affecting the DPAGT1 model.

Around 20% of all proteins are glycosylated [119], and membrane proteins (the targets of most drugs) are almost all glycosylated [120]. Given the role of DPAGT1 in N-glycosylation [6], it is possible that the glycosylation status of drug targets could bias our hits. Many of our suppressor drugs target proteins that are N-glycosylated. This includes acetylcholinesterase [121], DA receptors [122,123], DAT [124], NKCC1 [101], and COX-1/COX-2 [125]. However, it also includes acetylcholine receptors [38], which would be activated by acetylcholinesterase inhibitors and ostensibly be beneficial for the DPAGT1 model. In addition, the validated enhancer, ranitidine, targets the H2 receptor which has a putative N-glycosylation site (via UniProt [100]). Finally, three suppressors did not validate, but also target N-glycosylated receptors: this includes melatonin receptors [126], NMDA receptors [127], and adrenergic receptors [128,129] (S2 Table). While it is possible that the N-glycosylation status of a particular targeted protein is important for a specific drug hit, we found no general correlation between N-glycosylation status and suppressor or enhancer hits.

In this study, we screened 1,520 small molecules and identified multiple drugs that improved a Drosophila model of DPAGT1-CDG. We verify these findings using pharmacologic and genetic manipulation which strongly align with the known mechanisms of these therapeutic hits. We establish new biological connections between DPAGT1 and DA signaling, NKCC1, and prostaglandin synthesis. These findings may help create new treatment options for DPAGT1-CDG, and our validated drug hits represent potential therapeutics for patients.

Materials and methods

Fly stocks and maintenance

All flies were maintained at room temperature. All experiments were performed in a 20°C incubator unless otherwise noted. Flies were fed standard Glucose medium (D2) from Archon Scientific (Durham, North Carolina). We used fly stocks from the Bloomington Drosophila Stock Center and the Vienna Drosophila Resource Center [130] (listed in S3 Table). The DPAGT1 model (eya composite-GAL4, UAS-Alg7 RNAi [III]) was described previously and used the BDSC DPAGT1 RNAi stock 53264 (TRiP.GLC01825) [28]. Its genetic background control, eya composite-GAL4 (III), was a gift from Justin Kumar (Indiana University Bloomington). As described previously, we do not use the common GMR-GAL4 eye driver, as it is expressed later in development and did not result in a rough eye phenotype [28].

The second model of DPAGT1 deficiency (S1 Fig) was generated using BDSC DPAGT1 RNAi stock 51869. This stock was crossed to the eya composite-GAL4 (III) driver and balanced using the same Chr. III balancer, [TM3, Dfd-YFP, Sb], as the original DPAGT1 model. This resulted in w-, y-, v-, sev-; P[y+, v+, Alg7 RNAi]); (eya composite-GAL4, w+/TM3, Dfd-YFP, Sb) flies.

The DPAGT1 model containing the Dop1R1 null mutant (BDSC 92640) recombination (DPAGT1 model [Dop1R1^+/-]) was generated as follows. We crossed the DPAGT1 model [28] to the Dop1R1 null mutant (BDSC 92640) to create (eya composite-GAL4, w+, P[sc+, y+, v+, Alg7 RNAi])/(Dop1R1 null) (III) animals. We then crossed this line to a balancer line containing D/(TM3, ser) (III) and examined progeny for crossover events. We collected progeny with smaller eye size than the DPAGT1 model and ser. We then replaced the (TM3, ser) balancer with (TM3, Sb), self-crossed the line to ensure stability of the phenotype, and refer to this line as DPAGT1 model [Dop1R1^+/-] flies (eya composite-GAL4, w+, P[sc+, y+, v+, Alg7 RNAi], Dop1R1 null)/(TM3, Sb) (III).

RNA processing and qPCR sequencing

We collected 8–10 heads of 2–7 day-old female Drosophila of the eya composite-GAL4 background strain and the DPAGT1 model. We placed heads directly into 100 μL TRIzol reagent in eppendorf tubes, manually dounced the heads for 20 strokes using a plastic douncer, then placed these tubes onto ice. We then stored these samples for at least 24hrs at -80oC. We processed the TRIzol samples into RNA using the Zymo Direct-zol RNA Miniprep kit (Zymo Research cat. R2061), and we included the DNAseI treatment step. We converted RNA into cDNA using the ProtoScript II First Strand cDNA Synthesis Kit (NEB cat. E6560L). We then used this cDNA, PowerUp SYBR Green Master Mix (ThermoFisher cat. A25741), and forward/reverse primers to perform qPCR analysis on the QuantStudio 3 (ThermoFisher cat. A28567). Primers used: DPAGT1 F: ACTTCATGCTGCCTTTCCTG, DPAGT1 R: AAGTCATGCCGGCAAAGTAG; RpL19 F: AGGTCGGACTGCTTAGTGACC, RpL19 R: CGCAAGCTTATCAAGGATGG.

in vivo small molecule screen

For the primary screen, we used the Prestwick Chemical Library (PCL, Illkirch, France). The PCL contains 19 plates of 80 compounds each (1,520 total). All compounds came dissolved in DMSO at a concentration of 10 mM. Compounds were diluted further to 1 mM in phosphate-buffered saline. To make food containing these compounds, we used 500cc bags of standard Glucose medium (D2) from Archon Scientific (Durham, North Carolina). We dispensed this media into a beaker, microwaved until it liquified, then allowed it to cool on a heated stir plate under agitation. Once reaching 60°C, 1 mL of this media was dispensed into vials containing aliquots of the dissolved drugs or DMSO to reach a final concentration of 5 μM and 0.05% DMSO. We used up to eight DMSO control vials for each plate of 80 unique drugs.

Once the food was cooled, we placed 3–5 premated DPAGT1 model intercrossed females and 2–3 males into each vial. Flies were allowed to lay eggs for 1–4 days until visual inspection indicated that approximately 30 eggs were laid. Flies were then removed. The DPAGT1 model is balanced by TM3, Sb [28], and homozygotes are infrequent and semi-lethal. As such, we only collected heterozygote flies by selecting for the Sb phenotype. We collected up to five (average = 4.4, males and females), 2–7 day old progeny flies from each vial and froze them down at -80°C for later imaging. Drug names were masked, so both preparation of vials and collection of flies was done blinded to each drug.

Progeny eyes were imaged at 3x magnification (Leica EC3 camera). We determined eye area as previously described [131]. We masked image file names to blind observers to each treatment used. Within each set of 80 vials, we compared drug-treated animal eye sizes to the average of up to 8 DMSO controls and determined Z-scores (S1 Table). In one plate, drug-treated animals were instead compared to the plate average due to a technical error in the control (plate 03, S1 Table), but these values were in line with the rest of the plates. To determine if a drug was a hit, we used a Z-score threshold of 1.5 (derived from other Drosophila screens [132–134]).

We excluded all vials with a positive Z-score that contained only one fly (13 total vials, 0.9% of all 1,520 vials). 15 compounds were excluded for technical reasons (e.g. incorrect food pours, 1%). An additional 16 vials had eggs laid without any eclosed flies (1.1%). If ≤1 males were observed, but females were present, females were measured and compared to DMSO-treated females. This occurred in 47 vials (3.1%), and none of these reached our Z-score threshold. All reported Z-scores were derived from a single sex, and we indicate when females were scored in S1 Table.

Drug and RNAi validation

For making drug validation food, we used the same method as the primary screen, except that we used 10 mL of media, and we used drug concentrations of 1 μM, 5 μM, and 25 μM (as done previously [25]). We chose this more "conservative" dose curve as these drugs were already hits at 5 μM. Because there may be slight differences in drug concentrations from vendors or during food preparation, we considered drugs to validate if any of the three concentrations improved eye size (after multiple comparison correction). Most compounds were dissolved in DMSO. However, we employed other solvents if needed to reach the maximum concentration of 25 μM (S2 Table). We validated using at least two compounds from each drug category (S2 Table) from the Prestwick Chemical Library by using secondary vendors to limit quality control errors. These vendors were either Cayman Chemical (Ann Arbor, MI), MedChemExpress (Monmouth Junction, NJ), or Sigma-Aldrich (St. Louis, MO). Specific compounds and their vehicles are listed in S2 Table.

We crossed each fly knockout, overexpression, or RNAi line to either the DPAGT1 model or its control, w-;;eya composite-GAL4 (S3 Table). All RNAi in this study is driven by the eya composite-GAL4 line. RNAi or overexpression crosses used either attP40 (BDSC 36304), attP2 (BDSC 36303), or attP (VDRC 60100) as control comparisons, and null crosses used w¹¹¹⁸ (VDRC 60000). Because the DPAGT1 model uses RNAi endogenously, RNAi crosses resulted in double knockdown flies. When fly stocks were available, we used two different RNAi lines to reduce potential reagent-specific effects. Complete information on lines used can be found in S3 Table. The average N across every drug validation experiment was 14.5 flies (based on 3,186 fly measurements).

Both drug and RNAi validation used the same methodology as the primary screen for collecting animals and measuring eye sizes. Percentages listed in the text and S2 and S3 Tables represent the percent change of the treated mean compared to the control mean. See S4 Table for these raw data.

Statistics

For group comparisons, data were analyzed by One-way ANOVA with Welch’s correction and the post hoc Dunnett’s test to account for multiple comparisons. For individual comparisons, data were analyzed by Student’s t test with Welch’s correction. We used GraphPad Prism v10 or Microsoft Excel for these analyses.

Supporting information

S1 Table. List of all drugs tested in drug repurposing screen.

Each plate (first ##, 01–19) was tested independently of each other plate. See Methods for the Z-score calculation. "Females measured" means there were not enough males to measure, so females were measured instead. Excluded note meanings: "N = 1" means there was only one male (and one or no female) to measure; "No flies observed" means we did not observe any eclosed flies; "Technical reasons" means something unrelated to the drug negatively impacted the vial, such as an incorrect food pour.

(XLSX)

S2 Table. Information on drug validation experiments.

Tab information: "Primary hits" lists all drugs that reached at least 1.5 or -1.5 from the drug repurposing screen. "Rescuing drugs" lists all drugs that improved the DPAGT1 model (whether directly or derived from the screen). This includes the quantitative results of what % they improved eye size over vehicle-treated DPAGT1 model flies and their p-value (see Methods for statistics). The low/medium/high doses can be found in the tab "Compound sources, vehicles". "Non-rescuing drugs" lists drugs that worsened the model or had no effect. "Categories of tested drugs" lists the general drug class that each tested drug falls into. " Compound sources, vehicles" lists each drug tested during validation, where they were sourced from (including product numbers), their vehicle used, and doses used for each drug. Cells are labeled blue to indicate statistically significant positive changes, red represents negative changes, and gray represents non-significant changes.

(XLSX)

S3 Table. Information on gene validation experiments.

Quantitative results on the effect of genetic manipulations used on the DPAGT1 model, eya composite-GAL4 control, or DPAGT1 model [Dop1R1+/-]. This includes the specific stock number, the Drosophila gene affected, the nature of that stock (RNAi = RNA interference, OE = overexpression), the % change in eye size in each replicate, and the associated p-value (see Methods for statistics). Cells are labeled blue to indicate statistically significant positive changes, red represents negative changes, and gray represents non-significant changes. Each tab separates female and male experiments.

(XLSX)

S4 Table. Drug and RNAi validation raw data.

This file contains the raw pixel values of eye sizes of experimental and control conditions/genotypes for the validation experiments.

(XLSX)

S1 Fig. Knockdown in the DPAGT1 model and images of a second knockdown model.

(A) Graph of DPAGT1 knockdown in the DPAGT1 model, * p<0.05, (Student’s t-test). (B) Representative images of a second DPAGT1 knockdown model using the BDSC 51869 stock.

(PDF)

S2 Fig. Representative images of female fly stocks.

This is a complementary figure to Fig 1C to show what female eyes look like in each stock. Note that there is no image of the top "suppressor" or "enhancer" as the repurposing screen was done primarily in males.

(PDF)

S3 Fig. Representative images of drug-treated flies.

This includes both male and female flies and is complementary to each drug graph displayed here.

(PDF)

S4 Fig. Representative images of genetically manipulated flies.

This includes both male and females flies and is complementary to S3 Table.

(PDF)

Data Availability

All relevant data are within the manuscript and its Supporting Information files.

Funding Statement

CYC was supported by a NIGMS R35 award (R35GM124780) and a grant from the Primary Children's Hospital Center for Personalized Medicine, Salt Lake City, UT. HMD was supported by an NIGMS NRSA award (F32GM136057) and NICHD Career Transition Award (K99HD111662). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.