Synaptic roles for phosphomannomutase type 2 in a new Drosophila congenital disorder of glycosylation disease model

ABSTRACT Congenital disorders of glycosylation (CDGs) constitute a rapidly growing family of human diseases resulting from heritable mutations in genes driving the production and modification of glycoproteins. The resulting symptomatic hypoglycosylation causes multisystemic defects that include severe neurological impairments, revealing a particularly critical requirement for tightly regulated glycosylation in the nervous system. The most common CDG, CDG-Ia (PMM2-CDG), arises from phosphomannomutase type 2 (PMM2) mutations. Here, we report the generation and characterization of the first Drosophila CDG-Ia model. CRISPR-generated pmm2-null Drosophila mutants display severely disrupted glycosylation and early lethality, whereas RNAi-targeted knockdown of neuronal PMM2 results in a strong shift in the abundance of pauci-mannose glycan, progressive incoordination and later lethality, closely paralleling human CDG-Ia symptoms of shortened lifespan, movement impairments and defective neural development. Analyses of the well-characterized Drosophila neuromuscular junction (NMJ) reveal synaptic glycosylation loss accompanied by defects in both structural architecture and functional neurotransmission. NMJ synaptogenesis is driven by intercellular signals that traverse an extracellular synaptomatrix and are co-regulated by glycosylation and matrix metalloproteinases (MMPs). Specifically, trans-synaptic signaling by the Wnt protein Wingless (Wg) depends on the heparan sulfate proteoglycan (HSPG) co-receptor Dally-like protein (Dlp), which is regulated by synaptic MMP activity. Loss of synaptic MMP2, Wg ligand, Dlp co-receptor and downstream trans-synaptic signaling occurs with PMM2 knockdown. Taken together, this Drosophila CDG disease model provides a new avenue for the dissection of cellular and molecular mechanisms underlying neurological impairments and is a means by which to discover and test novel therapeutic treatment strategies. Drosophila Collection: This work generates a new Drosophila congenital disorder of glycosylation model for the most common disease category, caused by phosphomannomutase-2 mutation, and reveals a synaptic mechanism underlying associated neurological impairments.


INTRODUCTION
Congenital disorders of glycosylation (CDGs), which are caused by mutation of genes encoding glycosylation pathway proteins, are classified into two categories (Freeze et al., 2015): CDG-I disease states include defects in carbohydrate production, lipid-linked oligosaccharide (LLO) formation and attachment of glycan chains to amino acids; CDG-II disease states include defects in the modification and/or maturation of glycan chains after protein attachment. The most common CDG is a CDG-I called CDG-Ia or PMM2-CDG and results from mutations in phosphomannomutase 2 (PMM2), which converts mannose-6-phosphate to mannose-1phosphate, the obligatory precursor for GDP-mannose production and N-linked glycosylation (Andreotti et al., 2014;Freeze et al., 2014Freeze et al., , 2015. Since the first patient in 1980, >100 different mutations in >1000 CDG-Ia patients have been characterized (Freeze et al., 2014;Haeuptle and Hennet, 2009;Jaeken, 2013;Jaeken et al., 1980;Yuste-Checa et al., 2015). CDG-Ia infant mortality is ∼20% in the first year, with affected individuals manifesting subsequent increased susceptibility to organ failure, infection and injury (Grünewald, 2009). Individuals with CDG-Ia present with a spectrum of neurological symptoms (Jaeken, 2013), ranging from severe neurological impairments with early death, to mild defects with slight psychomotor delay (Grünewald, 2009;Marquardt and Denecke, 2003). To date, no effective treatments are available, with the only treatment option being symptom management (Grünewald, 2009;Monin et al., 2014;Stefanits et al., 2014).
CDG-Ia modeling is crucial for molecular and cellular studies. The initial mouse PMM2 knockout model has been of limited use owing to early embryonic lethality (Thiel et al., 2006). Heteroallelic combination of two PMM2 mutations allows partial enzymatic activity, protracted embryonic survival and has demonstrated potential maternal dietary intervention in treatment (Schneider et al., 2011). Increased lifespan occurred with mannose feeding prenatally and during gestation, reportedly allowing offspring to develop past critical periods of PMM2-dependent glycan requirement (Schneider et al., 2011;Thiel et al., 2006). Unfortunately, these results have thus far not been successfully replicated in patient trials, where postnatal oral and intravenous mannose administration failed to improve serum protein glycosylation levels (Kjaergaard et al., 1998;Mayatepek and Kohlmüller, 1998). A subsequent zebrafish model established via pmm2 morpholino knockdown revealed increased motor neuron number, altered cranial development, reduced motility and altered glycan profiles (Cline et al., 2012). Most recently, a similar Xenopus morpholino knockdown model demonstrated strong reduction in Wnt signaling, revealing a PMM2 requirement in intercellular communication (Himmelreich et al., 2015). These models have been valuable, but have limitations of morpholino-based approaches with inadequate targeting and concerns about the temporal maintenance of knockdown (Schulte-Merker and Stainier, 2014). We therefore set out to develop a Drosophila CDG-Ia disease model.
In this study, we generate a Drosophila CDG-Ia disease model. We show that Drosophila pmm2 is highly conserved, and we manipulate gene function by making mutants with clustered regularly-interspaced short palindromic repeat (CRISPR)/Cas9 genome editing and tissue-targeted transgenic RNAi. As in humans with CDG-Ia, Drosophila pmm2 loss-of-function (LOF) mutants exhibit reduced lifespan and psychomotor retardation proportional to the degree of PMM2 reduction. We show striking impacts on the N-linked glycome: globally in null mutants, neurally in tissue-targeted RNAi knockdown and locally at the NMJ synapse. At the NMJ, targeted pre-and post-synaptic PMM2 knockdown reveals architectural overelaboration, and concurrent reduction on both sides of the synapse strongly increases neurotransmission strength. PMM2 loss strongly impairs the synaptic matrix metalloproteinase (MMP) pathway regulating the HSPG co-receptor Dally-like protein (Dlp) to modulate Wnt Wingless (Wg) trans-synaptic signaling Dear et al., 2016;Friedman et al., 2013). Consistently, PMM2 knockdown reduces Wg, Dlp and the Frizzled nuclear import (FNI) pathway. These results suggest that a PMM2-dependent extracellular proteinase mechanism modulates Wnt signaling during NMJ synaptogenesis, which underlies coordinated movement and maintained viability. This new Drosophila model should lead to novel therapeutic treatments for CDG-Ia.

Degree of Drosophila PMM2 loss determines lifespan duration
Human PMM2 contains eight exons, compared to the Drosophila CG10688 pmm2 single reading frame (Fig. 1A), making genetic manipulation and expression studies more amenable. Simplified site-directed CRISPR mutagenesis targeted at one exon reduces the possibility of truncated exons maintaining partial function. Human PMM2 contains 12 active sites that coordinately bind the substrate, mannose-6-phosphate, and convert it to mannose-1-phosphate (Andreotti et al., 2014;Silvaggi et al., 2006). Drosophila PMM2 shows 100% conservation of these 12 active sites (Fig. 1A, red) and identifies numerous additional regions of high (90-100%) conservation (Fig. 1A, green). Overall, Drosophila PMM2 displays 56% amino acid identity with human PMM2 (Fig. 1A). To characterize the role of PMM2, we first made null mutants using CRISPR/Cas9 genome editing directed at both the 5′ and 3′ ends of Drosophila pmm2 ( Fig. 1A; Gratz et al., 2013). A total of 15 mutations were produced, all verified with direct sequencing: eight frameshifts, four insertions, two deletions and one missense mutation. Two pmm2-null frameshift mutations were selected for subsequent behavioral, functional and molecular studies, hereafter referred to as pmm2 FS1 and pmm2 FS2 .
Null pmm2 mutants, heteroallelic combinations and overdeficiency mutants all show identical early larval lethality during the late 1st to early 2nd instar transition (Fig. 1B) 1B). This suggests a delicate PMM2 balance, with strong neural loss being detrimental but moderate neural loss extending lifespan. Owing to these lethality constraints, hereafter weaker RNAi 107619 is used for ubiquitous knockdown and stronger RNAi 42956 is used for tissue-targeted neural and muscle knockdown.
Neuronal PMM2 maintains normal posture and coordinated movement Neurological movement symptoms in individuals with CDG-Ia range from slight gait ataxia to severe cerebellar ataxia; most children are unable to walk unassisted, and most adults are wheel-chair bound (Barone et al., 2014(Barone et al., , 2015Marquardt and Denecke, 2003;Monin et al., 2014). Loss of Drosophila PMM2 results in similarly severe postural and movement phenotypes, largely due to nervous system involvement. Targeted neural elav-Gal4 knockdown results in a highly penetrant 'held-out wings' posture ( Fig. 1C), which is associated with defects in flight-muscle control of wing positioning (Müller et al., 2010;Zaffran et al., 1997). Neural pmm2 knockdown adults have severe ataxia, with profound incoordination, inability to walk in a directed fashion and complete inability to fly. Weak muscle-specific PMM2 knockdown does not yield ataxia. However, strong muscle-specific knockdown allows pupae to develop fully, but not a single pharate adult properly ecloses. A few animals partially eclose before dying, and animals mechanically freed from pupal cases do not move or survive. Based on these severe qualitative movement defects, we conducted a range of quantitative analyses (Fig. 1D).

PMM2 loss suppresses N-glycosylation and enhances glycan turnover
To assess N-linked glycoprotein glycosylation levels correlating with movement defects in pmm2-null mutants and upon neurally targeted RNAi knockdown, we next assayed glycome composition. Humans with CDG-Ia display altered glycosylation status, including reduced concanavalin A (ConA) binding but increased fucose-decorated glycoproteins (Van Dijk et al., 2001). In Drosophila, at all stages of development and in all tissues, the major N-linked glycoprotein constituent, glycans, include highmannose and pauci-mannose (≥three mannose residues) classes, whereas human glycosylation typically results in more branched and complexly decorated proteins (ten Hagen et al., 2009;. Complex hybrid and branched N-glycans are present in Drosophila, but at much lower relative abundances . Here, Drosophila N-linked glycans were analyzed by mass spectrometry (MS) throughout the larval body in pmm2-null mutants and with combined neuronal (elav-Gal4) and muscle (24B-Gal4) pmm2 RNAi, and in adult heads only with neurally targeted pmm2 RNAi (Fig. 2).
Loss of N-linked glycoprotein glycosylation in pmm2 FS1 mutants implies decreased efficiency in glycosylation initiation by the oligosaccharyltransferase complex (OST) in the endoplasmic reticulum (ER), and/or increased ER-associated protein degradation (ERAD)-mediated protein deglycosylation within the cytoplasm (Gao et al., 2011). Either mechanism should be reflected by increased abundance of free oligosaccharide (FOS), branched oligosaccharides unbound to lipid or protein, either owing to endogenous peptide-Nglycosidase F (PNGase) release or via lipid-linked precursor hydrolysis by the OST (Cline et al., 2012;Gao et al., 2011). Consistently, FOS abundance was strikingly increased in pmm2 FS1 mutants (Fig. 3A). FOS abundance was detected as products of PNGase and endo-N-acetylglucosaminidase digestion (N1; Fig. 3A, blue asterisks) or glycans possessing two reducing terminal GlcNAc residues (N2; Fig. 3A, red asterisks), with total ion mapping chromatograms filtered for loss of non-reducing terminal HexNAc residues. Both FOS classes were increased in pmm2 nulls compared to controls (Fig. 3A). Thus, glycomics and FOS assays show striking reduction in mature glycan abundance globally, and an altered glycomic repertoire in the nervous system, with PMM2 loss. To test for a cellular role, we next moved to the well-characterized NMJ that drives movement.

PMM2 restricts NMJ structural growth and synaptic bouton differentiation
Loss of synaptomatrix glycosylation has been shown to result in elevated structural elaboration of the Drosophila NMJ, as evidenced by increased branching and excess synaptic boutons (Jumbo-Lucioni et al., 2014;Parkinson et al., 2013). To examine NMJ architecture, we characterized wandering 3rd instar muscle 4 NMJs labeled for HRP and Discs large (DLG) to illuminate pre-and postsynaptic compartments, respectively (Fig. 4A). Synaptic branches (defined as HRP-positive processes with >two boutons), boutons (defined as DLG-positive synaptic varicosities >1 µm in diameter) and NMJ terminal area (based on DLG-positive labeling) were all quantified ( Fig. 4B-D). Cell-specific roles were tested with neuronal (elav)-and muscle (24B)-targeted Gal4 drivers in comparison to ubiquitous (UH1-Gal4) pmm2 knockdown. In all cases, loss of PMM2 resulted in clear NMJ overelaboration, with more branches and supernumerary type Is/Ib synaptic boutons, with ubiquitous RNAi resulting in the greatest level of unrestrained NMJ overgrowth (Fig. 4).
Coupled pre-and post-synaptic PMM2 function limits NMJ transmission strength Glycosylation has been shown to play key roles in NMJ functional differentiation and the determination of neurotransmission strength (Dani et al., , 2014Parkinson et al., 2013). The severely impaired coordinated locomotion and alterations in NMJ structure similarly suggest that PMM2 has roles in synaptic function. We tested neurotransmission in the two-electrode voltage-clamp (TEVC) recording configuration by stimulating the motor nerve with a glass suction electrode and measuring the evoked excitatory junctional current (EJC) from the voltageclamped muscle (Parkinson et al., 2013). To compare EJC transmission properties, ten consecutive stimulation recordings were made at 0.2 Hz, and then averaged to calculate the mean peak transmission amplitude. Cell-specific roles in functional differentiation were tested with targeted neuronal (elav) and muscle (24B) Gal4 drivers, alone and in combination, in comparison to ubiquitous (UH1-Gal4) pmm2 RNAi knockdown (Fig. 6).
PMM2 positively regulates the synaptic extracellular matrix proteinase pathway Extracellular mechanisms in the highly glycosylated NMJ synaptomatrix provide an obvious answer to the PMM2 non-cell-autonomous phenotype. We were first guided to consider extracellular MMP pathways owing to the common tracheal break and melanization mutant phenotypes (Glasheen et al., 2010;Page-McCaw et al., 2007;Zhang and Ward, 2009) shared with PMM2 LOF (data not shown). Subsequently, recent work has shown that mmp mutants exhibit both NMJ structural and functional phenotypes that are strikingly similar to PMM2 LOF (Dear et al., 2016). Therefore, we examined the matrix metalloproteome at the NMJ, which includes secreted MMP1, glycosylphosphatidylinositol (GPI)-anchored MMP2 and their shared secreted tissue inhibitor of MMP (TIMP) (Dear et al., 2016;Kessenbrock et al., 2010;Page-McCaw et al., 2007). We hypothesized that PMM2-dependent glycan modification of these extracellular proteins, and/or their synaptic substrates, could provide a mechanism regulating NMJ structure and function.

PMM2 positively regulates the Wg trans-synaptic signaling pathway
MMPs play an important role in Wg intercellular signaling by directly regulating the Wg co-receptor, the HSPG Dlp (Wang and Page-McCaw, 2014). Importantly, the same Wg-Dlp signaling pathway is a critical driver of structural and functional development at the Drosophila NMJ (Ataman et al., 2006(Ataman et al., , 2008Kerr et al., 2014;Mathew et al., 2005) and is known to be modulated by glycan mechanisms Jumbo-Lucioni et al., 2014;Parkinson et al., 2013). Wg binds the Frizzled-2 (Fz2) receptor, which is internalized and the C-terminus proteolytically cleaved (Fz2C) for transport via the Fz2C nuclear import (FNI) pathway (Ataman et al., 2006;Mathew et al., 2005) to modulate NMJ structure and/or function (Speese et al., 2012). Based on this extensive work, we hypothesized that PMM2 regulates MMP2- dependent Wg signaling to regulate NMJ structure and function, which underlies coordinated movement and maintained viability.

DISCUSSION
We set forth to establish a Drosophila CDG-Ia (aka PMM2-CDG) genetic model through manipulation of the causative PMM2 gene (Freeze et al., 2014). Using CRISPR-generated pmm2-null mutants and transgenic RNAi, we found that PMM2 levels correlated to coordinated movement abilities and lifespan (Fig. 1), as in individuals with CDG-Ia (Cylwik et al., 2013;Jaeken, 2013;de Lonlay et al., 2001). Humans with identical PMM2 mutations present with a wide spectrum of movement defects (Marquardt and Denecke, 2003;Schneider et al., 2011), attributed to genetic and/or environmental factors that can be readily controlled in Drosophila. With tissuespecific drivers, we found a neural PMM2 impairment for coordinated movement (Fig. 1). Interestingly, weak neural knockdown of pmm2 resulted in increased lifespan. Similarly, moderate impairments of oxidative-stress and dietary-restriction pathways have also been reported to extend lifespan (Mair et al., 2005;Min and Tatar, 2006;Ristow and Schmeisser, 2011). Like PMM2, severe impairments of these pathways result in reduced lifespan and early death, but more modest impairments extend lifespan via changes in metabolic rate, developmental conditioning and/or defense mechanisms. Null pmm2 mutants displayed severe attenuation of glycoprotein glycosylation (Fig. 2), with reduced N-linked glycosylation diversity . Similar global N-glycan losses occurred with strong pmm2 RNAi throughout the larval neuromusculature, but not with weaker neural-targeted pmm2 RNAi in the adult head (Fig. 2). Lipid-linked oligosaccharides (LLOs) used for protein attachment by OST activity are regulated at many levels (Gao et al., 2011). PMM2 loss should reduce LLO levels by inhibiting mannose-1-phosphate production and elevating the mannose-6-phosphate pool, which acts as a signal mediating LLO destruction, thereby increasing FOS levels ( Fig. 3A; Gao et al., 2011). However, the increase in pauci-mannose structures with adult neuronal pmm2 knockdown flies differentially alters glycan maturation. The mannose-6-phosphate increase might change the mannose phosphoisomerase (MPI) equilibrium, leading to interconversion of mannose-6-phosphate to fructose-6-phosphate to siphon LLO-toxic mannose-6-phosphate and mitigate LLO elimination.
Null pmm2 mutants displayed elevated FOS levels (Fig. 3A). Increased phosphorylated FOS levels, predicted to be cleaved from LLO intermediates, likewise occur in CDG-Ia patient cells (Vleugels et al., 2011). Similarly, the zebrafish morpholino model shows increased FOS levels, with phenotype rescue via MPI coreduction, suggesting causative mannose-6-phosphate elevation (Cline et al., 2012). Viable human CDG-Ia patients are typically heterozygous for pmm2 mutations, resulting in partial loss of PMM2 (Matthijs et al., 1998;Monin et al., 2014). In Drosophila, partial LOF from neuron-targeted RNAi resulted in surprising resistance to glycosylation changes (Fig. 2C). Pauci-mannose glycans were increased, but high mannose and complex glycans unchanged. PMM2 regulates GDP-mannose availability for glycan production, but, following glycosylation by OST, further processing should be GDP-mannose independent (Cylwik et al., 2013). The increased pauci-mannose glycans resulting from PMM2 partial loss suggests that GDP-mannose levels influence glycan processing beyond the role as synthetic donors for mannosylation. Paucimannose glycan production is driven by the balance between Golgi exomannosidases, GlcNAc-transferase 1 and a hexosaminidase removing GlcNAc from nascent complex glycans . Precursor abundances, including of donor and acceptor, influence expression and activity of glycan-processing enzymes. Reduced GDP-mannose likely skews the balance of enzyme activities that trim high-mannose to pauci-mannose, before they increase in complexity. Impacts on glycosylation independent of GDP-mannose were also evident by the observation of reduced NMJ lectin labeling with PMM2 loss (Fig. 3B). Importantly, both ECL and VVA bind mannose-free structures, most likely on O-linked glycoprotein backbones.
The nervous system is tightly regulated by glycans at multiple levels of development and function (Koles et al., 2007;Scott and Panin, 2014a,b;Seppo and Tiemeyer, 2000). NMJ architectural overelaboration and functional strengthening has been identified in many glycan mutants in our systematic genetic screens, including sulf1, mgat1, galt and pgant (Dani et al., , 2014Jumbo-Lucioni et al., 2014;Parkinson et al., 2013). Consistently, PMM2 loss showed striking NMJ overelaboration (Fig. 4), supporting the conclusion that glycans primarily inhibit synaptic morphogenesis. Null pmm2 mutants already exhibited strong synaptic architecture defects within a day after hatching (Fig. 5), showing that PMM2dependent glycan mechanisms brake the earliest stages of synaptic growth and differentiation. Glycosylation mechanisms also have key roles in modulating neurotransmission (Koles et al., 2007;Scott and Panin, 2014a,b), and PMM2 loss strongly increased NMJ function (Fig. 6). There was no effect on spontaneous synaptic vesicle release or post-synaptic amplitude, indicating a specific PMM2 role in limiting stimulus-evoked QC. Interestingly, neuraland muscle-targeted PMM2 knockdown had no effect on transmission; however, combined neural and muscle knockdown fully replicated the ubiquitous PMM2-loss phenotype (Fig. 6), indicating that the elevated transmission needs concomitant PMM2 removal both pre-and post-synaptically. One explanation is that semi-synaptic glycosylation might be sufficient to normalize transmission: loss of glycosylation from one side might be compensated for by the other side, because synapse transmission is highly regulated by both pre-and post-synaptic cells . Another idea is that glycosylation could be provided by extracellular components from either synaptic partner cell.
The non-cell-autonomous defect occurring with PMM2 loss prompted us to investigate extracellular signaling mechanisms, which are tightly regulated by glycosylation at the NMJ (Dani et al., , 2014Parkinson et al., 2013;Jumbo-Lucioni et al., 2014;Rushton et al., 2012). In particular, MMPs play crucial roles shaping synapse structure and function (Dear et al., 2016;Kessenbrock et al., 2010;Page-McCaw et al., 2007;Sternlicht and Werb, 2001). PMM2 loss could impair MMP glycosylation or its ability to cleave improperly glycosylated substrates (Godenschwege et al., 2000;Llano et al., 2000Llano et al., , 2002Pohar et al., 1999). PNGaseF and EndoH treatment to remove N-linked glycans showed that MMP2 isoforms are highly glycosylated, whereas only a tiny subset of MMP1 isoforms are glycosylated (Fig. 7A). GPI-anchored MMP2 also requires GDP-mannose, although GPI anchors require fewer donor mannose than N-linked glycans (de la Morena-Barrio et al., 2013). Importantly, pmm2 RNAi similarly removed glycosylation from multiple MMP2 isoforms, with only a minor change to MMP1 (Fig. 7A). As predicted by PMM2-dependent glycosylation changes, MMP2 levels were strongly reduced at the NMJ synapse with pmm2 RNAi, whereas those of MMP1 were not significantly altered (Fig. 7B,C). The TIMP regulator was also reduced in abundance with removal of PMM2, which would be predicted to help alleviate consequences of MMP2 loss, perhaps as a compensation mechanism (Dear et al., 2016). Importantly, recent work from our lab has shown that MMPs play crucial roles regulating NMJ structural and functional synaptogenesis via the control of HSPG receptors that modulate Wnt trans-synaptic signaling (Dear et al., 2016).
Recent work utilizing PMM2 morpholino knockdown in Xenopus revealed altered glycosylation of Wingless-type MMTV integration site family growth factor (Wnt) and reduction of Wnt signaling (Himmelreich et al., 2015). Similarly, we find that pmm2 RNAi knockdown in Drosophila suppresses Wnt Wingless (Wg) signaling at the developing NMJ synapse (Fig. 8). With PMM2 loss, synaptic levels of Wg ligand and its HSPG co-receptor Dlp were both strongly reduced, and downstream signaling through the FNI (Speese et al., 2012) pathway was consistently downregulated (Fig. 8). These defects have been previously associated with the loss of synaptic MMP2, which acts via Dlp to regulate Wg trans-synaptic signaling to modulate both structural and functional NMJ development in the same direction (Dear et al., 2016). However, we have shown that PMM2 loss has myriad consequences on N-linked glycoprotein glycosylation, and therefore quite likely impacts NMJ synaptogenesis at multiple levels. Indeed, specifically targeted reduction in Wg signaling alone has previously been associated with decreased NMJ structural development and reduced function (Ataman et al., 2008;Kerr et al., 2014;Packard et al., 2002), which differs from the Wg attenuation associated with PMM2 loss. Therefore, PMM2 roles at the NMJ synapse likely reflect roles in multiple intersecting pathways that jointly control growth, structural differentiation and neurotransmission strength. Our future work will be aimed at deciphering other PMM2-dependent glycoprotein contributions, which combinatorially result in the structural and functional NMJ defects characterizing this CDG-Ia disease state model.
We hope that this new Drosophila model will prove instrumental for tackling the disease, and related CDGs, especially in regard to neurological symptoms (Grünewald, 2009;Jaeken, 2010). One avenue will be to dissect roles played by glycan precursors and mannose-6-phosphate buildup, by examining genetic interactions shifting the relative abundance of alternatively processed glycans and to alleviate increased FOS levels; for example, by using genetic MPI reduction or pharmaceutical MPI inhibitors in the benzoisothiazolone series (Sharma et al., 2011). One such agent, MLS0315771, has been shown to favor mannose-1-phosphate production in CDG-Ia patient fibroblasts and zebrafish embryos. Pharmacological tests in Drosophila could include assays to prolong lifespan, improve coordinated movement, and prevent NMJ structural and functional defects. The current standard of care for CDG-Ia patients is simply symptomatic treatment and disease management (Grünewald, 2009;Jaeken, 2013). The mouse model suggests beneficial dietary intervention, common for other metabolic disorders like classic galactosemia (Jumbo-Lucioni et al., 2014). However, mannose treatment has not been effective in restoring N-linked glycoprotein glycosylation levels in CDG-Ia patients (Thiel and Körner, 2013). Drug avenues to increase PMM2dependent glycosylation are hypothesized, but there are no studies (Thiel and Körner, 2013). We expect that the relatively high speed of Drosophila disease model studies utilizing the powerful Drosophila genetic toolkit will open up new avenues for disease intervention. We propose here that targeting the matrix metalloproteome and Wnt signaling pathways offers potential new candidates to consider in developing future CDG-Ia treatments.

PCR methods
For reverse transcription quantitative PCR (RT-qPCR), total RNA was extracted using a Zymo Research Direct-zol RNA Miniprep Plus Kit with TRI reagent (R2070) with on-column DNase treatment. The Superscript VILO cDNA synthesis kit (11754-050) was used for cDNA synthesis. RT-qPCR was run on a Bio-Rad CFX96 with equal amounts of cDNA (2 ng for each trial). For expression quantification, the Pfaffl method was used with standards of known transcript number to quantify absolute cDNA number for target and reference genes. For the reference, ribosomal protein L32 (CG7939) levels were used for normalization of absolute cDNA quantity [(target/ reference)×100]. The following primers were used for target and reference: pmm2 forward 5′-AGGCTCGGATCTGGAGAAGA-3′, pmm2 reverse 5′-AATGTCGTACTCGGCGAACA-3′; L32 forward 5′-CGGTTACGGAT-CGAACAAGC-3′, L32 reverse 5′-CTTGCGCTTCTTGGAGGAGA-3′. Samples and standards were run with gene-specific primers in duplicate trials. n=5 tissue collection samples and n=15 assay replicates. 50 first instars were used for RNA extraction.

Behavioral assays
Egg lays were collected overnight on apple juice agar plates. Plates were then cleared of larvae, and newly hatched larvae collected after 1 h (t=0). Larval lifespan analyses involved daily counts. Adult lifespan analyses required two separate methods. For strong neural elav-Gal4>RNAi pmm2 knockdown and elav-Gal4/+ controls, adults were collected at eclosion and maintained in laying pots with filter paper covering apple juice plates with yeast. For all other lifespan assays, adults were maintained in normal fly tubes on cornmeal/agar/ molasses food. Adult survival was measured 3 times/week, with animals transferred to fresh plates or tubes. The comparative quantification of survival is reported as the time to which 50% of the animals remain viable [half-time survival (HTS)]. Adult and larval locomotion were assayed as previously described (Nichols et al., 2012;Sokolowski, 1980). Briefly, larvae were tested on apple juice agar plates with yeast paste spread around edges as an attractant. Individual larvae were placed in the middle of a plate and time-lapse-recorded under a dissection microscope with a Canon Rebel DSLR camera (Melville, NY). Locomotion was assayed as peristaltic waves per second (w/s) (Gjorgjieva et al., 2013). Adult motility was assayed by negative geotaxis and a ring locomotion assay. For geotaxis, adults were placed in empty fly vials for 15 min to acclimate, and then tubes were sharply tapped to put animals at the bottom (Nichols et al., 2012). Movement was recorded with a Canon Rebel DSLR camera and the percentage of animals to climb above 2 cm measured at timed intervals. To assay horizontal locomotion, a 4-cm circle was drawn in a large dish, and flies with amputated wings were placed in the middle of the circle. Movement was recorded with a Canon Rebel DSLR camera and the time required to traverse the circle measured.

Glycomic analyses
Glycoproteins and free oligosaccharides (FOS) were prepared from staged collections of pmm2-null 1st instars or elav-Gal4>RNAi 42956 adult heads by homogenization in aqueous/organic solvents and subsequent protein precipitation as described previously . Briefly, the aqueous/ organic homogenate was centrifuged with glycoproteins recovered in the pellet and FOS from the supernatant . Precipitated proteins were washed with cold acetone and dried under a stream of nitrogen to produce samples stored desiccated at −20°C. Protein content was determined by bicinchoninic acid (BCA) assay of resolubilized material (Pierce). Nlinked glycoprotein glycans were prepared from 1-mg aliquots by digestion with trypsin/chymotrypsin, followed by enzymatic release of glycans with PNGaseF . FOS were separated by passage over a Sep-pak C 18 cartridge column . N-linked glycans and FOS were permethylated and analyzed by mass spectrometry (MS) using nanospray ionization coupled to linear iontrap and orbital Fourier transform mass analyzers (Discover NSI-LTQ/OrbitrapFT, Thermo-Fisher Scientific). MS spectra were collected over the range m/z=200-2000 and MS/MS fragmentation by collision-induced dissociation (CID; 30-40% normalized collision energy) was acquired over the same m/z range using the total ion mapping function of the XCalibur instrument software (version 2.0). Annotated glycans were validated by exact mass in full MS and by manual inspection of MS/MS spectra at each of the detected m/z values.

Immunocytochemistry imaging
Immunocytochemistry studies were performed as described previously (Parkinson et al., 2013). Briefly, all animals were dissected, fixed and labeled identically in the same dish. Wandering 3rd or 1st instars were dissected in physiological saline containing 128 mM NaCl, 2 mM KCl, 4 mM MgCl 2 , 0.25 mM CaCl 2 , 70 mM sucrose, 5 mM trehalose and 5 mM HEPES ( pH 7.1). Preparations were fixed in 4% paraformaldehyde for 10 min at room temperature (RT) in phosphate-buffered saline (PBS). Preparations were then either processed with detergent [PBS+1% bovine serum albumin (BSA)+0.2% Triton X-100] for intracellular labeling, or detergent-free (PBS with 1% BSA) for extracellular studies. . Primary antibodies and lectins were incubated at 4°C overnight; secondary antibodies were incubated at RT for 2 h. Samples were mounted in Fluoromount-G (Electron Microscopy Sciences, Hatfield, PA). All preparations were imaged using identical parameters. Z-stacks were taken with a Zeiss LSM510 META laser-scanning confocal using a 63× Plan Apo oil-immersion objective. Optical sections were imaged starting above and ending below the NMJ or muscle nuclei to encompass their entirety. Stacks were projected on the Z-axis for maximum intensity, with NMJ or nuclei signals highlighted and average intensity quantified using ImageJ (Abramoff et al., 2001).

Electrophysiology
Excitatory junctional current (EJC) recordings made using two-electrode voltage-clamp (TEVC) were done as previously reported (Parkinson et al., 2013). Briefly, wandering 3rd instars were glued with 3M Vetbond adhesive (World Precision Instruments, Sarasota, FL) to sylgard-coated glass coverslips, cut longitudinally along the dorsal midline, internal organs removed and sides affixed down for neuromusculature access. Peripheral nerves were cut at the ventral nerve cord. Recordings were done at 18°C in saline consisting of 128 mM NaCl, 2 mM KCl, 4 mM MgCl 2 , 1 mM CaCl 2 , 70 mM sucrose, 5 mM trehalose and 5 mM HEPES ( pH 7.1), imaged using a Zeiss Axioskop microscope with 40× immersion objective. A fire-polished glass suction electrode was used for evoked nerve stimulation with 0.5 ms suprathreshold stimuli at 0.2 Hz from a Grass S88 stimulator (Rohrbough et al., 2007). Muscle 6 in abdominal segments 2/3 was impaled with two microelectrodes of 10-15 MΩ resistance filled with 3 M KCl, and clamped (−60 mV) using an Axoclamp-2B amplifier (Molecular Devices, Sunnyvale, CA). EJC records were filtered at 2 kHz. To quantify EJC amplitudes, ten consecutive traces were averaged. Spontaneous miniature EJC (mEJC) records were made in 2-min sessions and filtered at 200 Hz with a low-pass Gaussian filter prior to quantification. Clampex software was used for all data acquisition, and Clampfit software for all data analyses (Molecular Devices, Sunnyvale, CA).

Statistics
All statistical analyses were performed using GraphPad InStat3 software (La Jolla, CA). Student's t-tests were used for pairwise comparisons, and ANOVA with appropriate post-hoc testing was used for all data sets of three or more comparisons. Nonparametric methods were used for data sets lacking normal distribution. Fisher's exact tests were used to analyze contingency tables for adult behavioral data. Data are shown as mean±s.e.m. in all figures, with significance presented as P≤0.05 (*), P≤0.01 (**), P≤0.001 (***) and P≤0.0001 (****).
This article is part of a subject collection on Spotlight on Drosophila: Translational Impact. See related articles in this collection at http://jcs.biologists.org/collection/ drosophila-disease-model.