Overelaborated synaptic architecture and reduced synaptomatrix glycosylation in a Drosophila classic galactosemia disease model

Classic galactosemia (CG) is an autosomal recessive disorder resulting from loss of galactose-1-phosphate uridyltransferase (GALT), which catalyzes conversion of galactose-1-phosphate and uridine diphosphate (UDP)-glucose to glucose-1-phosphate and UDP-galactose, immediately upstream of UDP–N-acetylgalactosamine and UDP–N-acetylglucosamine synthesis. These four UDP-sugars are essential donors for driving the synthesis of glycoproteins and glycolipids, which heavily decorate cell surfaces and extracellular spaces. In addition to acute, potentially lethal neonatal symptoms, maturing individuals with CG develop striking neurodevelopmental, motor and cognitive impairments. Previous studies suggest that neurological symptoms are associated with glycosylation defects, with CG recently being described as a congenital disorder of glycosylation (CDG), showing defects in both N- and O-linked glycans. Here, we characterize behavioral traits, synaptic development and glycosylated synaptomatrix formation in a GALT-deficient Drosophila disease model. Loss of Drosophila GALT (dGALT) greatly impairs coordinated movement and results in structural overelaboration and architectural abnormalities at the neuromuscular junction (NMJ). Dietary galactose and mutation of galactokinase (dGALK) or UDP-glucose dehydrogenase (sugarless) genes are identified, respectively, as critical environmental and genetic modifiers of behavioral and cellular defects. Assaying the NMJ extracellular synaptomatrix with a broad panel of lectin probes reveals profound alterations in dGALT mutants, including depletion of galactosyl, N-acetylgalactosamine and fucosylated horseradish peroxidase (HRP) moieties, which are differentially corrected by dGALK co-removal and sugarless overexpression. Synaptogenesis relies on trans-synaptic signals modulated by this synaptomatrix carbohydrate environment, and dGALT-null NMJs display striking changes in heparan sulfate proteoglycan (HSPG) co-receptor and Wnt ligand levels, which are also corrected by dGALK co-removal and sugarless overexpression. These results reveal synaptomatrix glycosylation losses, altered trans-synaptic signaling pathway components, defective synaptogenesis and impaired coordinated movement in a CG neurological disease model.


INTRODUCTION
Classic galactosemia (CG; OMIM 230400) results from loss of galactose-1-phosphate uridyltransferase (GALT), the second enzyme in the Leloir pathway, which acts immediately downstream of galactokinase (GALK), the initial enzyme (McCorvie and Timson, 2011). GALT maintains the balance between uridine diphosphate (UDP)-glucose (glc), -galactose (gal), -N-acetylgalactosamine (GalNac) and -N-acetylglucosamine (GlcNac) (Frey, 1996). Levels of these four UDP-sugars are ratelimiting for the biosynthesis of glycoproteins and proteoglycans (Freeze and Elbein, 2009), which form the foundation of the extracellular synaptomatrix of the synaptic cleft and perisynaptic space . UDP-glc dehydrogenase is encoded by Drosophila sugarless (sgl) (Häcker et al., 1997;Toyoda et al., 2000), and sgl mutations compromise biosynthesis of the heparan sulfate proteoglycan (HSPG) co-receptor Dally-like protein (Dlp), which is known to regulate trans-synaptic signaling of the Wnt protein Wingless (Wg); such signaling drives neuromuscular junction (NMJ) synaptogenesis . These studies implicate a core pathway involving GALT, GALK and Sgl in the regulation of HSPG co-receptor control of Wnt signaling during NMJ synapse formation, and indicate that disruption of this pathway is a potential causal mechanism underlying CG neuropathology.
Acute CG neonatal symptoms are alleviated by dietary galactose restriction (Jumbo-Lucioni et al., 2012), but maturing individuals with CG develop substantial neurodevelopmental, motor and cognitive impairments (Ridel et al., 2005). After >50 years of research, there is still no mechanistic understanding of these chronic neurological symptoms. However, a long-term and extensive body of studies documents glycosylation defects in individuals with CG (Haberland et al., 1971;Petry et al., 1991;Charlwood et al., 1998;Liu et al., 2012). Galactose is a major component of complex carbohydrates in glycoproteins and glycolipids in the nervous system, and defective glycosylation impairs neurodevelopment and neurological function . In particular, the heavily glycosylated NMJ synaptomatrix plays crucial roles in synaptogenesis during normal development, and its disruption is implicated in numerous heritable disease states . For example, glycosylation defects are causal in numerous muscular dystrophies (MDs) and congenital disorders of glycosylation (CDGs) that are characterized by severe neurological impairments (Muntoni et al., 2008;Freeze, 2013).
We recently conducted a Drosophila screen of glycogenes via RNAi knockdown of N/O-linked glycans, glycosaminoglycans, glycosyltransferases and glycan-binding lectins to test the effects on NMJ structure and function . This screen identified Drosophila GALT (dGALT) as a potent regulator of NMJ architecture. We therefore set forth to characterize synaptic morphology and glycosylated synaptomatrix composition in the recently established Drosophila CG disease model (dGALT Overelaborated synaptic architecture and reduced synaptomatrix glycosylation in a Drosophila classic galactosemia disease model Patricia Jumbo-Lucioni, William Parkinson and Kendal Broadie* deficiency) (Kushner et al., 2010), and to identify glycan mechanisms driving synaptogenic defects. We found that dGALT nulls exhibit a profoundly altered carbohydrate landscape within the NMJ synaptomatrix, accompanied by loss of the HSPG co-receptor Dlp and extracellular accumulation of Wg ligand. Crucially, synaptomatrix defects were differentially corrected by dGALK coremoval and sgl overexpression. Consistently, we found that dGALK removal and sgl overexpression in dGALT mutants corrected both the motor defects and NMJ architectural abnormalities resulting from dGALT loss of function. We conclude that dGALT, dGALK and sgl define a genetic pathway regulating synaptomatrix glycosylation state to modulate components of a Wnt trans-synaptic signaling pathway, and thereby control NMJ synaptic morphogenesis to support coordinated movement.

Impaired coordinated movement of dGALT nulls is rescued by human GALT expression
Evidence amassed over decades from individuals with CG documents common movement defects (Waggoner et al., 1990;Schweitzer et al., 1993;Kaufman et al., 1995;Hughes et al., 2009;Rubio-Agusti et al., 2013). Similarly, dGALT-null mutants display defects in startle-induced, geo-negative climbing behavior (Ryan et al., 2012). To assay locomotion defects directly, we first assayed daily motor activity levels in individual animals using the Drosophila Activity Monitoring (DAM) system, which measures movement disruption of an infrared beam (Chiu et al., 2010). Adult animals aged 3-5 days were compared between dGALT ΔAP2 (null) and dGALT C2 (precise-excision genetic control) flies. Both genotypes were entrained to 12:12-hour light:dark cycles for 2 days, and then activity counts were recorded for ≥3 days. dGALTnull mutants were significantly (P<0.01) movement-impaired compared with controls (20.2±1.8 versus 27.7±2.1 counts/hour during the light period; n≥20 for each genotype). These results show that removal of dGALT strongly reduces locomotor activity in this CG model.
We next turned to the well-characterized movement behaviors in Drosophila larvae (Heckscher et al., 2012). Bilateral coordinated movement can be quantified with a rollover assay (Bodily et al., 2001;Pan et al., 2008). We tested wandering L3 larvae that had complete loss of dGALT (dGALT ΔAP2 , n=37 and dGALT ΔAP2 /Df, n=34), and compared with a genetic control (dGALT C2 , n=53). Compared with the righting time of controls (17.20±1.03 s), both dGALT (30.95±2.59 s; P<0.001) and dGALT/Df (27.81±3.04 s; P<0.05) animals were significantly slower and obviously less coordinated (Fig. 1B). [Note that, in all graphs shown in the figures, fold change compared with control values (set at 1) are shown.] There was no significant difference between the two null conditions. To confirm that the movement defect is mediated by dGALT loss, we next drove wild-type human GALT (UAS-hGALT) in the dGALT ΔAP2 -null background ( Fig. 1B; 'rescue', n=14). Transgenic hGALT expression in the dGALT ΔAP2 -null background completely rescued normal coordinated movement compared with the driveralone control (dGALT; UH1-Gal4/+, n=15; P=0.006), which was indistinguishable from homozygous dGALT ΔAP2 mutants (dGALT; Fig. 1B). These results demonstrate that dGALT activity is necessary for proper coordinated movement, and that this requirement is functionally conserved with human GALT. Because irregularities in NMJ architecture can cause locomotor impairments, we next characterized synaptic structure under conditions of intact, complete loss of and partial loss of dGALT activity.

Loss of dGALT causes striking structural defects at the NMJ synapse
Our recent Drosophila RNAi screen identified that dGALT is required in NMJ morphological synaptogenesis . The glutamatergic NMJ, which drives coordinated movement, has been well characterized in Drosophila larvae (Andlauer and Sigrist, 2012;Collins and DiAntonio, 2007), and the structure and function of this synaptic terminal is highly dependent on appropriate expression of extracellular glycans . We therefore hypothesized that defects in NMJ development underlie motor impairments in dGALT mutants Courchesne et al., 2011). Axonal growth properties, branch formation and synaptic bouton differentiation together shape the complex threedimensional synaptic architecture (Broadie et al., 2011;Nahm et al., 2013). NMJs from wandering L3 larvae were labeled with anti-

Clinical issue
Classic galactosemia (CG) results from complete or almost complete loss of galactose-1-phosphate uridyltransferase (GALT), the second enzyme in the Leloir pathway of galactose metabolism. GALT catalyzes the generation of a glucose precursor and maintains the balance between uridine diphosphate (UDP) sugars, the obligate sugar donors for the synthesis of glycoproteins and glycolipids (glycosylation), which are important components of cell membranes and extracellular matrix. Dietary galactose restriction alleviates neonatal lethal symptoms of CG, but affected individuals still develop substantial neurological complications (including motor and cognitive impairments) of unknown etiology. CG has recently been classified as a congenital disorder of glycosylation (CDG), and glycosylation defects have been suggested as the underlying mechanism for the chronic neurological disease symptoms. Beyond the single CG condition, more than 60 other CDGs have been identified so far. Neurological impairments are common in these CDG disease states, but effective treatments are not yet available.

Results
A recent Drosophila screen of glycosylation-related genes identified GALT as a potent regulator of the structure of the neuromuscular junction (NMJ; the synapse connecting motor neurons and muscle fibers). To extend this discovery, GALT-deficient Drosophila (dGALT nulls; a CG model) have been used to characterize movement behavior, NMJ synapse structure and function, and the carbohydrate composition of the synapse extracellular environment (i.e. synaptomatrix) in the disease.
The results show that loss of GALT activity impairs coordinated movement, causes structural synapse overelaboration, profoundly changes the carbohydrate composition of the NMJ synaptomatrix, and alters the Wnt trans-synaptic co-receptor and ligand abundance (known to control synaptic morphogenesis at the NMJ). In double mutants, the combination of GALT deficiency either with loss of galactokinase (GALK; an enzyme that acts upstream of GALT) or overexpression of UDPglucose dehydrogenase (an enzyme that acts downstream of GALT; encoded by sugarless) corrects the glycosylation defects, the NMJ architecture alterations and the movement impairments associated with GALT loss in this Drosophila CG model.
Implications and future directions NMJ defects likely underlie the pathogenesis of movement disorders frequently reported in CG-affected individuals, and might also account for similar impairments that characterize other CDG disease states. The findings reported here are the first to reveal NMJ synaptic glycan loss, synaptic architecture defects and trans-synaptic Wnt co-receptor and ligand defects as potential causal factors for the well-known CG-related movement disabilities. Furthermore, the results presented here suggest that targeting galactokinase and UDP-glucose dehydrogenase could modify CG outcome and represent a potential therapeutic approach in this disease. Future investigations are needed to advance the translation of findings on CG-related candidate genes into the development of novel drugs, and to extend this approach to the study of other CDG diseases.
To test NMJ neurotransmission strength, the motor nerve was stimulated with a glass suction electrode while recording from the voltage-clamped muscle . Excitatory junction current (EJC) recordings were made at suprathreshold stimulation levels at a 0.5 Hz frequency. Four genotypes were compared: genetic background control (dGALT C2 ; precise excision) to homozygous-null mutant (dGALT ΔAP2 ; imprecise excision), and UH1-Gal4-driven UAS-dGALT-RNAi to UH1-Gal4/+-alone transgenic control. More than ten NMJs from ≥five different animals were recorded for each of the four Although there was a trend towards increased neurotransmission in both comparisons, neither change was significant. We therefore conclude that dGALT does not significantly alter neurotransmission strength at the NMJ, but rather plays a specific role in controlling synaptic architecture, which we therefore focused on in subsequent studies. Because exposure to dietary galactose is a well-recognized modifier of acute outcome in CG (Jumbo-Lucioni et al., 2013), we next characterized locomotion and NMJ structural defects with galactose dietary supplementation.
It has been previously shown that a high-galactose diet fed to wildtype larvae causes accumulation of galactose-1-phosphate (gal-1-P) (Kushner et al., 2010), which was previously shown to alter the UDPsugar balance (Lai et al., 2003). To further test whether motor behavior and NMJ structural defects occurring in galactose-fed conditions are linked to gal-1-P accumulation, dGALK nulls (dGALK ΔEXC9 ), with undetectable enzymatic activity, were raised either on food supplemented with 200 mM galactose or on normal food. Galactosefed mutants took as much time as unsupplemented mutants to rollover from an inverted to an upright position (8.84±0.90 s versus 9.5±1.57 s; n≥8). Similarly, both NMJ branch number and interbouton distances were not significantly different in the two conditions (data not shown). Animals displayed 2.0±0.19 (n=8) and 2.2±0.17 (n=15) branches under galactose-free and galactose-fed conditions, respectively. Interbouton distance was 0.81±0.08 μm (n=15) and 0.77±0.06 μm (n=8) with and without galactose feeding, respectively. However, galactose-fed dGALK mutants displayed excess NMJ boutons compared with unsupplemented animals (23.27±1.23 versus 14.57±1.02; n≥8, P<0.0001) and significantly greater NMJ cumulative length (113.08±6.04 μm versus 81.65±4.36 μm, n≥8, P=0.002). Because a high-galactose diet can alter glycosylation status, and abnormal glycosylation of the extracellular environment at the NMJ is well known to modulate synaptogenesis , we next characterized the glycosylated synaptomatrix composition of the NMJ.
dGALT activity shapes the glycosylated synaptomatrix composition of the NMJ Glycosylation defects in plasma and tissue samples have long been reported in individuals with CG (Haberland et al., 1971;Jaeken et al., 1992;Sturiale et al., 2005). These glycan errors persist even after prolonged dietary galactose restriction, and have therefore been suggested to account for chronic neurological symptoms in such individuals (Charlwood et al., 1998). Importantly, abnormally glycosylated synaptomatrix at the Drosophila NMJ is well known to alter synapse formation , and therefore provides a probable mechanistic basis for the synaptic architecture defects manifested in dGALT-null mutants (Figs 1, 2). To test this hypothesis, we next assayed the NMJ synaptomatrix with a panel of lectins labeling (1)  Representative images and a summary of the major changes are shown in Fig. 3.
Because control animals fed galactose replicate dGALT phenotypes, we next explored whether galactose feeding in wild-type larvae alters the NMJ glycosylation composition, by assaying WFA lectin expression with and without dietary intervention. WFA lectin strongly labeled unsupplemented control NMJs (normalized 1.0±0.05, n=8) but, similar to dGALT mutants, was strikingly reduced in larvae raised on a high-galactose diet (0.78±0.07; n=7, P=0.02). To further test the role of dGALT as a modifier of NMJ glycosylation state, transgenic hGALT rescue animals were compared with two control groups (wild type, and driver-alone in the dGALT ΔAP2 background) by assaying WFA expression. The loss of WFA lectin signal in dGALT nulls (0.72±0.07; n=18) was completely rescued with transgenic expression of hGALT (1.08±0.12; n=7) to a level undistinguishable from wildtype controls (1.0±0.07; n=19). Taken together, these results reveal striking alterations in NMJ glycan composition in the absence of dGALT activity, specifically including reductions in galactosyl and Nacetylgalactosamine residues, and fucosylated HRP glycans. Because the NMJ synaptomatrix carbohydrate environment limits synaptic morphogenesis via the modulation of trans-synaptic signaling of the Wnt Wg, the pathway that drives morphological synaptogenesis Parkinson et al., 2013), we next characterized components of this signaling pathway at the NMJ.
dGALT mutants display changes in Wnt trans-synaptic signaling pathway components Glycan mechanisms within the heavily glycosylated Drosophila NMJ synaptomatrix have been repeatedly shown to regulate synaptic morphogenesis . In particular, glycans modulate the trans-synaptic signaling that drives synaptogenesis Friedman et al., 2013;Parkinson et al., 2013). More specifically, the HSPG Dlp has been identified as a mediator of synapse formation Van Vactor et al., 2006). At the Drosophila NMJ, Dlp acts as a co-receptor of the Wnt protein Wg to limit extracellular Wg availability and presentation, regulating NMJ development Friedman et al., 2013). Importantly, this mechanism is linked to the Leloir pathway via UDP-glc dehydrogenase (Fig. 1A), whose loss reduces Dlp biosynthesis (Binari et al., 1997;Toyoda et al., 2000). We therefore hypothesized that loss of dGALT would alter Dlp expression to misregulate Wg ligand abundance, with the change in Wg trans-synaptic signaling providing a mechanistic basis for NMJ synaptogenesis defects in the CG disease model. To test this hypothesis, we probed NMJs from wandering L3 larvae with anti-Dlp and -Wg, using anti-HRP as the synaptic marker. Data and representative images are shown in Fig. 4.
To test the hypothesis that dGALK co-removal would prevent the losses in glycosylation of the NMJ synaptomatrix observed in dGALT nulls, we next probed double mutants with lectins for the three glycans with the largest changes in levels observed in the single mutants (Fig. 3): ECL lectin for terminal galactosyl residues, WFA lectin for N-acetylgalactosamine residues and anti-HRP for fucosylated HRP epitopes (Fig. 6). The losses of ECL and WFA lectin labeling in dGALT nulls completely disappeared with dGALK co-removal (Fig. 6A,B). ECL lectin labeling was once again reduced by ~50% in dGALT nulls (P<0.001), but dGALK co-removal restored ECL labeling to control levels (n=19; Fig. 6A′). Loss of dGALT activity similarly decreased WFA labeling by ~50% (n=17) compared with controls (n=23, P<0.001), but WFA labeling was again fully restored to control levels in double mutants (n=10; Fig. 6B′). Loss of GALK alone impacted neither ECL nor WFA labeling (n=6 for both). In contrast, the reduced HRP signal observed in dGALT-null NMJs was unaffected by eliminating dGALK in parallel: in the absence of dGALT (0.74±0.06, n=22), there was a highly significant reduction in HRP labeling in mutants compared with controls (1.0±0.03, n=36, P<0.001), but there was no significant improvement with dGALK co-removal (0.81±0.05, n=25). Thus, the double mutant selectively restores synaptomatrix glycosylation state, but does not completely reverse NMJ glycan changes.
In light of such glycosylation state changes, we next tested whether dGALK co-removal also restores Dlp and Wg levels (Fig. 7). Normalized to controls, reduced Dlp levels characterizing dGALT nulls (0.64±0.05, n≥8, P<0.001) were significantly elevated towards wild-type values in double mutants (0.87±0.13, n≥9, P<0.01) and were no longer significantly different from controls ( Fig. 7A,C). Likewise, elevated Wg levels in dGALT-null NMJs (1.45±0.14, n≥8, P<0.001) were reduced towards wild-type levels in dGALT; dGALK double nulls (1.25±0.10, n≥9), and were no longer significantly different from controls (Fig. 7B,D). Thus, the reduced levels of Dlp co-receptor and increased levels of Wg ligand at dGALT-null NMJs were significantly corrected by co-removal of  dGALK activity. Taken together, these results demonstrate that dGALK is a key genetic modifier in the Drosophila CG disease model, raising the intriguing possibility that inhibition of GALK activity in patients might similarly modulate CG neurological symptoms. Because sgl is essential for Dlp biosynthesis (Häcker et al., 1997), we next tested whether increasing Dlp biosynthesis via sgl overexpression could similarly prevent defective NMJ synaptogenesis and impaired movement.

DISCUSSION
Loss of motor coordination has long been reported as one of the most frequent CG symptoms (Böhles et al., 1986;Waggoner et al., 1990;Schweitzer et al., 1993;Kaufman et al., 1995;Hughes et al., 2009). Consistently, the Drosophila CG disease model showed reduced activity and impaired movement coordination, which were fully rescued by transgenic expression of human GALT, demonstrating full functional conservation. The mechanistic basis of CG movement defects remains unknown after 50 years of study

RESEARCH ARTICLE
Disease Models & Mechanisms (2014) doi:10.1242/dmm.017137  (Bosch, 2006). However, our recent Drosophila screen of glycogenes identified a role of dGALT in NMJ morphological synaptogenesis . In light of previous reports in rodent (Audouard et al., 2012) and Drosophila (Feiguin et al., 2009) providing compelling evidence associating deficient movement behaviors with NMJ structural abnormalities, we set forth to characterize NMJ synaptic architecture under conditions of complete and targeted loss of dGALT activity. These studies reveal elevated synaptic growth and structural overelaboration, albeit without a change in basal NMJ transmission strength. These findings are consistent with previous studies showing that rollover movement defects occur independently of synaptic transmission defects (Bodily et al., 2001), and suggest that defective NMJ architecture impairs the muscle control that is required to coordinate the sequence of motor outputs to optimally produce a bilateral coordinated movement (twist-and-roll behavior). These NMJ structural defects are also consistent with our earlier work showing that loss of specific N-glycans , or the whole cassette of complex/branched N-glycans , removes constraints on NMJ growth and structural elaboration. We conclude that glycans play a primarily inhibitory role in modulating synapse morphogenesis.
Major controversies surround the benefits of a galactoserestricted diet in the CG disease state (Jumbo-Lucioni et al., 2012), particularly in alleviating chronic neurological symptoms (Bosch, 2006). We therefore tested impacts of galactose in the diet on movement and NMJ structure in our Drosophila model. A highgalactose diet did not impact either the impaired movement or overelaborated NMJ architecture of dGALT mutants, consistent with earlier reports that dGALT long-term phenotypes are independent of dietary galactose (Ryan et al., 2012). Strikingly, however, wild-type animals fed a high-galactose diet phenocopied many dGALT-mutant phenotypes, displaying slowed coordinated movement and grossly overelaborated NMJs with supernumerary boutons over decreased interbouton distances. Previous studies on experimental models of galactosemia in genetically wild-type animals (Cui et al., 2006;Wei et al., 2005;Long et al., 2007) demonstrate that exposure to high levels of galactose leads to neurodegeneration and cognitive disability, some of the chronic complications reported in CG. Increased dietary galactose decreases the UDP-glc:UDP-gal ratio (Gibson et al., 1995;Gibson et al., 1996). Because glycosylation depends upon UDP-sugar availability, NMJ architectural defects in both wild type overfed galactose and dGALT nulls might result from disrupted UDP-sugar balance and consequently impaired glycosylation. Drosophila larvae have been shown to accumulate significantly higher levels of gal-1-P on a high-galactose diet (Kushner et al., 2010), which has been shown to directly impact UDP-sugar balance (Lai et al., 2003). In contrast to the Drosophila CG model, the viability of mice deficient in GALT activity and reared on a high-galactose diet is reportedly unaffected, despite the accumulation of very high gal-1-P levels (Leslie et al., 1996;Ning et al., 2001). A theory to explain this discrepancy is that very low levels of aldose reductase activity in mice prevent the accumulation of galactitol, a toxic galactose intermediate. A newly established GALT-null mouse model (Tang et al., 2014) reveals reduced viability and abnormal cellular changes in the brain as a result of galactose exposure, but fails to inform whether such changes persist under galactose restriction. Our findings in galactose-fed dGALK mutants suggest that gal-1-P accumulation is a primary determinant of the behavioral deficits characterizing dGALT mutants, and wild-type animals fed a high galactose diet.
Glycan-binding lectins have long been used to define the extracellular glycan landscape at the NMJ synapse (Ohtsubo and Marth, 2006;Martin and Freeze, 2003;Scott et al., 1988). Null dGALT NMJs displayed striking glycosylation defects, including substantial reductions in galactosyl, N-acetylgalactosamine and fucosylated HRP moieties. Differences between lectin probes arise from different binding specificity. ECL and PNA lectins both bind terminal galactose, whereas PNA binds preferentially Gal(β-1,3)-GalNAc and ECL binds Gal(β1,4)-GlcNAc. Similarly, WFA recognizes terminating N-acetylgalactosamine (α/β-linked to galactose) and VVA binds preferentially to Tn-antigen (i.e. a single α-GalNAc residue linked to serine or threonine). Earlier studies examining the molecular basis of neural anti-HRP staining in Drosophila have demonstrated specific recognition of core α1,3fucosylated glycoproteins (Fabini et al., 2001). Galactose-containing glycans have key synaptogenesis roles (Tai and Zipser, 1999), and glycosylation losses in rodents (Lowe and Marth, 2003) and specific loss of HRP epitopes in Drosophila (Baas et al., 2011) have both been independently linked to motility abnormalities. However, although other CDG models also display evidence of overelaborated motor neuron architectures (Cline et al., 2012), the correlation with loss of HRP epitopes is less clear. Two previous studies have shown a reduction of HRP expression with underelaborated Drosophila NMJ structure (Rendić et al., 2010;Baas et al., 2011), in contrast to our results, whereas a recent study from our lab demonstrated HRP loss accompanied by increased synaptic growth and structural overelaboration , as in the current study. Such complicated carbohydrate-mediated tuning of NMJ synaptogenesis might explain the correction of the synaptic architectural overelaboration in dGALT nulls with dGALK coremoval or sgl overexpression, despite the differential and complementary correction of synaptomatrix glycosylation losses.
NMJ synaptogenesis requires bidirectional trans-synaptic signaling via secreted glycoprotein ligands (Carbonetto and Lindenbaum, 1995;Collins and DiAntonio, 2007). Such signals must necessarily traverse the heavily glycosylated synaptomatrix, and we have shown that the glycan state of this extracellular environment is crucially important for enabling and shaping trans-synaptic signaling during NMJ synaptogenesis (Rohrbough and Broadie, 2010;Rushton et al., 2012;. This synaptomatrix glycosylation state is strongly compromised in our CG disease model, suggesting that signaling should be similarly altered. Consistently, dGALT mutants exhibited elevated Wnt Wg ligand, the best-characterized trans-synaptic signal at the Drosophila NMJ (Kamimura et al., 2013;Miech et al., 2008) The HSPG Dlp acts as a Wg co-receptor, regulating both extracellular distribution and signaling (Han et al., 2005;Kirkpatrick et al., 2004), and we have recently established that Dlp limits Wg trans-synaptic signaling at the Drosophila NMJ Friedman et al., 2013). Consistently, dGALT nulls exhibited a sharp decrease in Dlp levels, providing a mechanism for Wg overexpression. Wg overexpression in turn is well known to increase NMJ bouton formation (Packard et al., 2002;Ataman et al., 2008), providing a mechanism to explain the supernumerary boutons characterizing our CG disease model. Importantly, defects in trans-synaptic Wg co-receptor and ligand levels are rescued with transgenic expression of hGALT, showing functional conservation.
Sgl is required for the synthesis of the HSPG co-receptor Dlp (Toyoda et al., 2000;Haerry et al., 1997), which modulates Wg signaling, as described above. Consistently, sgl mutants recapitulate Wg phenotypes (Superina et al., 2014;Haerry et al., 1997). This established interaction strongly supports findings here showing that sgl is a strong genetic modifier in our CG disease model. Moreover, sgl overexpression by itself phenocopies loss of dGALT, and sgl gain-of-function in dGALT mutants restores NMJ architecture, Wg trans-synaptic signaling components and motor behavior output. In wild type, increased UDP-glc dehydrogenase activity can augment hyaluronan production, glycosaminoglycan release and extracellular matrix elaboration (Clarkin et al., 2011), and thus impact glycosylated synaptomatrix composition to account for negative outcomes of sgl overexpression. Conversely, overproduction of extracellular glycans is predicted to be beneficial under conditions of a depleted glycosylated synaptomatrix, as occurs in dGALT nulls. Furthermore, Dlp and Wg levels are both strongly corrected toward control values by dGALK co-removal and sgl overexpression genetic interventions. These findings strengthen the argument that the Wnt signaling pathway plays a crucial role in the pathogenesis of neurological complications in CG. Taken together, these results suggest that both GALK inhibition and UDP-glc dehydrogenase activation might combat neurological symptoms in individuals with CG.
In conclusion, the results presented here are the first to reveal NMJ glycosylation losses, synaptic architecture defects and concomitant changes in trans-synaptic Wnt co-receptor and ligand levels in a CG disease model. These findings suggest a model in which loss of GALT activity triggers gal-1-P accumulation, which subsequently limits UDP-sugar bioavailability. In this model, UDPsugar deficits trigger changes in NMJ glycosylated synaptomatrix composition, including levels of GPI-anchored HSPG Dlp, an essential co-receptor for a Wnt ligand during NMJ synaptogenesis. Loss of HSPG regulation results in failure to control Wg levels, causing differential trans-synaptic Wnt signaling to promote excess growth and overelaborated architectural complexity. Co-removal of dGALK, and sgl overexpression, both are strong genetic modifiers of outcome in this CG model. We propose that similar defects underlie well-characterized glycosylation defects and movement disorders in the human CG disease state, and might account for neurological pathogenesis characterizing a wide array of related CDG disease states, which will be the subject of our future investigations. Fig. 1A shows the Leloir pathway, listing enzyme names and Drosophila CG numbers of genes targeted in this study. A dGALT imprecise-excision null, dGALT ΔAP2 , with undetectable enzyme activity, and a precise-excision control, dGALT C2 , with normal enzyme activity, were previously described (Kushner et al., 2010). The genomic deficiency Df(2L)BSC187 crossed to dGALT ΔAP2 was used as a second heterozygous-null condition. Ubiquitous (UH1) and tissue-specific (neuronal elav and muscle 24B) Gal4 drivers were used to drive dGALT-RNAi (v10025; P{KK102974}VIE-260B, Vienna Drosophila RNAi Center). Gal4 drivers alone (Gal4/+) crossed to w 1118 were genetic controls. A dGALK imprecise-excision null, dGALK ΔEXC9 , with undetectable enzyme activity, was generated by mobilizing a P-element insertion (EY03791) in the 5′-UTR of CG5068. UAS-sgl UY771 was used to overexpress sgl (Monnier et al., 2002). dGALT, dGALK and sgl alleles were combined using standard genetic techniques to generate double mutants. Single and double mutants were reared at 25°C on standard molasses-based food. To test effects of a high-galactose diet, food was supplemented with 200 mM galactose.

Behavioral assays
All behavioral experiments were carried out on at least eight individual animals for each genotype. Locomotor activity was assayed in adult males using the Drosophila Activity Monitoring (DAM) system (Trikinetics, Waltham, MA). Starting 1 day after eclosion, animals were entrained to a 12-hour light/dark cycle for 2 days before data recording, with activity counts per hour recorded for 3 days (Chiu et al., 2010). Coordinated larval movement was assayed in male wandering third instars (L3) using the rollover assay described previously (Bodily et al., 2001;Pan et al., 2008). Animals were placed individually on a room temperature (RT) 1% agar plate and allowed to acclimate for 2 minutes. The L3 animal was then rolled to an inverted position as defined by the ventral midline. Once released, a timer was used to measure the time the animal took to completely right, as defined by the dorsal midline. Three consecutive measurements were recorded for each animal and averaged to produce one data point. Data were analyzed by Student's t-test for pairwise comparisons, and ANOVA tests for all data sets of ≥three comparisons.
All mutant and control larvae were dissected, labeled and imaged in parallel. z-stacks were taken with a Zeiss LSM 510 META confocal using 40/63× oil-immersion objectives, with sections starting immediately above and ending immediately below the NMJ. Expression analyses of conjugated lectins and anti-HRP were performed using NIH ImageJ software with the threshold function outlining lectin-or HRP-labeled NMJ within FasIIdefined synaptic regions. Anti-Dlp and anti-Wg expression analyses were quantified within HRP-defined synaptic regions. Fluorescence was normalized to control values from the same experiment with all imaging parameters kept constant between compared genotypes. For structural analyses, preparations were double-labeled with anti-HRP and anti-DLG, with counts made at muscle 4 in segments A2/3 on the right and left sides. Data were averaged for each animal or hemisegment to produce one data

RESEARCH ARTICLE
Disease Models & Mechanisms (2014) doi:10.1242/dmm.017137 point for expression and structural analyses, respectively. For structural quantification, a bouton was defined as an axon varicosity >1 μm in minimum diameter, and ≥two boutons on one axon defined an NMJ branch. NMJ cumulative length was measured by combining branch lengths, and inter-bouton distances obtained as an average from single branches with ≥six boutons.

Statistical analyses
Behavioral, functional, structural and fluorescence intensity data were averaged per genotype, calculated as a fold-change relative to the mean control value from the same experiment. Unpaired two-tailed t-tests with Welch correction were applied for normally distributed datasets with unequal standard deviations. For data not normally distributed, pairwise comparisons were done with non-parametric Mann-Whitney tests. One-way analysis of variance (ANOVA) was used for parametric multiple comparisons with Tukey-Kramer post-tests. For nonparametric multiple comparisons, Kruskal-Wallis tests were applied with Dunn's multiple comparisons. All statistical analyses were performed using GraphPad InStat version 3.0 (GraphPad Software). Significance in figures is presented as P<0.05 (*), P<0.01 (**), P<0.001 (***) and P<0.0001 (****).