Age-dependent degeneration of an identified adult leg motor neuron in a Drosophila SOD1 model of ALS

Amyotrophic lateral sclerosis (ALS) is an adult onset neurodegenerative disease characterized by the progressive loss of motor neurons (Boillee et al., 2006). Post-diagnosis survival times average 3–5 years and currently there is no treatment available that substantially extends life despite the availability of two therapeutic drugs, riluzole and edaravone (Sawada, 2017). Much of our understanding of disease mechanisms derives from the study of dominant, single gene mutations causing familial ALS (fALS), which represents 10% of the total disease population. To date, 16 genes have been identified with clear causality to fALS (Taylor et al., 2016). Mechanistically, these gene products are involved in a range of cellular processes including autophagy, RNA metabolism, cytoskeletal transport, oxidative stress response and inflammation (reviewed in Taylor et al., 2016 and Nguyen et al., 2018). Together these genes help to define pathways important for disease pathogenesis.

Over 150 mutations in superoxide dismutase (SOD1) cause fALS (http://alsod.iop.kcl.ac.uk/) (Forsberg et al., 2019; Wroe et al., 2008). To date, all mutations increase the potential of the normally dimeric SOD1 to form destabilized monomers and aggregates (Lang et al., 2015). SOD1 fibrillization kinetics show strong, inverse correlations with disease duration (Lang et al., 2015; McAlary et al., 2016). Transgenic mouse models expressing mutant human SOD1 (hSOD1), accumulate hSOD1 positive aggregates in the cytoplasm and mitochondria of motor neurons (MNs). These animals also show age-dependent locomotor impairment, progressive paralysis and MN degeneration. Specifically, fast-fatigable MNs are lost first by apoptotic or necroptotic cell death mechanisms involving bcl-2/bax or ripk1 pathways (Morrice et al., 2017).

Although SOD1 is expressed ubiquitously and abundantly, restricted expression of mutant hSOD1 in either muscle or neurons can reproduce MN loss in mouse models (Dobrowolny et al., 2008; Jaarsma et al., 2008; Wong and Martin, 2010). These data suggest that neuromuscular junction (NMJ) maintenance, which requires signaling between the neuron, glia and muscle at the tripartite synapse is important for disease initiation and progression. The ‘dying-back’ of axons is an early feature of disease and is observed at pre-symptomatic stages (Fischer et al., 2004). Also, denervation precedes accumulation of SOD1-positive aggregates and is associated with mitochondrial vacuolization (Gould et al., 2006). Taken together, these results indicate that changes at the NMJ are early and critical features of ALS pathogenesis.

Although the various SOD1 models of ALS reproduce many disease features, they rely on overexpression of mutant hSOD1 in a genetic background, which includes endogenous expression of wild-type sod in the model organism. In mice, multiple copies of mutant hSOD1 are required to generate disease-related phenotypes and overexpression of wild-type hSOD1 alleles produced disease-like phenotypes including altered mitochondrial morphology, motor neuron degeneration, ataxia and shortened life span (Graffmo et al., 2013; Jaarsma, 2006; Jaarsma et al., 2000). A fly overexpression model produced similar results in which wild-type hSOD1 or the mutant variants G85R or A4V were driven in MNs using the bipartite Gal4/UAS system (Watson et al., 2008). Although expression of mutant hSOD1 generally produced more severe phenotypes, aged D42-Gal4>UAS-hSOD1⁺ flies showed progressive locomotor defects and electrophysiological defects in the giant fiber pathway innervating the DLM and TTMs. These results raise the important question of the extent to which ALS model phenotypes are caused by an artifact of overexpression versus presence of disease-causing mutations.

Previously, we characterized a novel Drosophila ALS model created by a gene-replacement strategy (Sahin et al., 2017). Using ends-out homologous recombination, we replaced dsod1⁺ with ALS-associated mutations including G85R, H48R or H71Y. Each of these three alleles produced recessive phenotypes that ranged in severity from early adult lethality prior to eclosion for dsod1^G85R/G85R to shortened lifespan in dsod1^H71Y/H71Y mutants (Sahin et al., 2017). Interestingly, recessive ALS-like phenotypes are seen in canine and murine models in which endogenous sod is mutated, and also in a G85R-YFP transgenic mouse model (Wang et al., 2009; Joyce et al., 2015; Katz et al., 2017). As adults, dsod1^H71Y/H71Y homozygotes show shortened life span and progressive locomotor defects; however, there is no readily apparent MN loss based on TUNEL labeling studies or cell counts using nuclear-localized GFP expressed in glutaminergic neurons. Similar preservation of MNs is generally observed in G85R homozygotes; however, these flies die as pharate adults preventing the study of progressive changes in the adult central nervous system (CNS) (Sahin et al., 2017).

Although no MN loss was detected previously, we sought to determine whether there are distal changes in MN morphology over time in dsod1-mutant flies. Although the patterning, structure and function of the Drosophila NMJ is well established for identified MNs in larvae, the larval NMJ is maintained for only a short time before neuronal and muscle remodeling occurs after puparium formation (Brown et al., 2006; Watts et al., 2003). Therefore, the study of slow degenerative changes superimposed upon rapid development of larvae into pupae presents a challenge in well-characterized larval MNs. Alternatively, the study of adult neurons has focused mainly on non-motor neuron cell types including photoreceptor cells, CNS interneurons, olfactory or gustatory neurons or neurosecretory cell populations (for example, see Hussain et al., 2018; Logan et al., 2012). Recent work has begun characterizing adult MNs that innervate the abdomen, the proboscis or the leg (Hebbar and Fernandes, 2004; Hebbar et al., 2006; Mahoney et al., 2014). Degeneration of leg motor and sensory neuron populations have been studied in the context of neurodegeneration (Sreedharan et al., 2015). This work identified Shaggy/GSK3, Hat-trick and Xmas-2 as suppressors of TDP-43 induced toxicity using genetic mosaics to drive mutant TDP-43 in restricted cell populations. In addition, Fernius et al. (2017) used leg motor and sensory neurons to profile degenerative phenotypes for various neurodegenerative diseases created by overexpressing various mutant alleles (Fernius et al., 2017). These studies relied on imaging MNs through the cuticle using GFP expressing lines coupled with relevant gene mutations.

Here we characterize changes in an identified adult MN (MN- I2) in dissected adult legs using standard immunohistochemical techniques and thus eliminating the need for complex genetic backgrounds with fluorescent tags to label structures. The adult leg of Drosophila is comprised of approximately 50 MNs that innervate the musculature to drive coordinated walking behavior (Baek and Mann, 2009; Brierley et al., 2009, 2012; Enriquez et al., 2015). The leg MNs arise from 11 neuroblast lineages and produce stereotyped axonal and dendritic arborization patterns (Baek and Mann, 2009). Our results show that MN- I2 axon morphology is normal in newly eclosed dsod1^H71Y/H71Y homozygotes but undergoes progressive changes resulting in highly disorganized axonal projections with an overall reduction in axonal branch length and synaptic bouton number. In aged dsod1^H71Y/H71Y animals, we find ‘ghost’ branches and associated boutons that lack the postsynaptic marker DLG. These changes reproduce several key features of the dismantling of the NMJ seen in ALS. Furthermore, similar phenotypes seen in dsod1^null/null aged-adults suggest these changes result from SOD1 loss-of-function. Finally, this work establishes a model to study slow progressive changes in an identified MN over time.

Denervation is a prominent feature of ALS, characterized by reduction in motor units and synapses leading to muscle wasting and eventually paralysis. Changes at the NMJ can be observed at presymptomatic stages in animal models and therefore represent early events in disease pathogenesis (Hegedus et al., 2007, 2008; Martineau et al., 2018). In Drosophila genetic ALS models, presynaptic changes have been characterized for mutant human fused in sarcolemma (FUS) driven within neurons. These flies showed altered active zone organization and concomitant decreases in evoked post-synaptic currents (Shahidullah et al., 2013). In TAR DNA-binding protein-43 homolog (tbph) mutants, decreased locomotion, loss of presynaptic terminals, and loss of MN populations were observed when tbph was driven or knocked-down (Diaper et al., 2013). These changes were accompanied by reduced synaptic transmission. There were also decreases in the number of small boutons and impaired BMP signaling at the larval NMJ when human TBPH was driven in neurons (Deshpande et al., 2016). In contrast, Drosophila dsod1 mutants reveal only subtle changes in larval MNs (Held et al., 2019; Sahin et al., 2017). Because dsod1^G85R/G85R flies show pharate adult lethality with shortened legs and MNs with broken axonal branches, adult MNs appear to be preferentially affected in dsod1 mutants. To characterize adult MN phenotypes independent of development, we use longer-lived dsod1^null and dsod1^H71Y alleles, which live median life spans of 13.5 and 6.5 days, respectively, and animals develop normally (Sahin et al., 2017). Previously, we showed age-dependent progressive locomotor defects in dsod1^H71Y/H71Y flies (Sahin et al., 2017). Given age-dependent disease progression in ALS, we wondered whether there were neuromuscular changes that correlated with adult locomotor defects in aged dsod1^H71Y/H71Y flies.

We assessed adult walking phenotypes using footprinting assays as a measure of locomotor ability in dsod1 mutants and dsod1⁺ flies. The dsod1⁺ genotype used throughout this study contains one loxP site within the single intron of dsod1⁺ at an identical position found in mutant dsod1 stocks as a result of ends-out homologous recombination and removal of the mini-w⁺ selectable marker. The dsod1^null allele was generated by CRISPR and removed the entire ORF as diagramed in Fig. S1. Flies locomote using a repeating tripod or tetrapod gate (Mendes et al., 2013). We analyzed walking patterns for deviations from this repeating pattern and quantified the percentage of flies with either normal or abnormal gait patterns (Fig. 1). Patterns considered abnormal ranged from deviations in pattern spacing to pronounced leg dragging. In newly eclosed adults, >80% of dsod1^+/+ flies showed normal walking gaits and this frequency remained high though day 10 aged adults (Fig. 1B). We found lower percentages of flies with normal gaits beginning at eclosion for both mutant dsod1 alleles. In addition, both dsod1^H71Y/H71Y and dsod1^null/null mutants showed age-related declines in normal locomotion between days 0 to 6 for H71Y (67% to 22%), and days 6 to 10 for null (80% to 40%), respectively. Walking in aged dsod1^H71Y/H71Y was often associated with a prominent metathoracic leg dragging phenotype (as shown in Fig. 1A, dsod1^H71Y/H71Y day 6). Abnormal walking in dsod1^null/null was generally less severe with 0% displaying leg-dragging phenotypes on days 6 and 10 compared to 44% and 17% for dsod1^H71Y/H71Y. These results confirm and extend our initial published observations (Sahin et al., 2017).

The study of leg MNs is emerging as a model system for the study of slow neurodegenerative processes. In addition, the development of leg MNs through neuroblast lineage analysis and their arborization patterns has been well documented. These studies lay the foundation for further characterization of single neuron populations over time in adults. In ALS, MN populations are differentially affected with fast-fatigable MNs degenerating first followed by fast-fatigue resistant and then slow MNs (Martineau et al., 2018). Thus, there is utility in assessing phenotypes in individual MNs and we focus on the axonal projections from one of two neurons derived from the presumptive ‘I’ neuroblast lineage based on developmental analysis (Baek and Mann, 2009). Both MN-I1 and MN-I2 are glutaminergic neurons that innervate the tibia levitator muscle (TLM) of the femur (Fig. S2). Both MNs were readily identifiable with modest arborization patterns but MN-I2 was less frequently damaged upon dissection in our hands. Based on these practical considerations, we focused on MN-I2 for further analysis.

To first determine whether MN-I2 makes stereotyped patterns of innervation, we performed Sholl analysis of axon arbors using dsod1⁺ (loxP) control flies. Because some adult neurons show sexually dimorphic arborization patterns, we compared legs from male and female flies at eclosion (day 0) and representative images and tracings are shown in Fig. 2. While there was no difference in branching intersections between left and right legs, there was a small difference between males and females (P<0.05 for the right leg). We further quantified basic parameters of MN-I2 and found reproducible features across individuals including number of branches (7.1±0.7); number of boutons (83.7±9.8) and area of arborization (3898±1010).

Reproducible MN-I2 axonal branching allowed us to track changes in innervation patterns over time and we sought to determine whether aspects of dying-back were apparent in the metathoracic legs of ALS-associated dsod1^H71Y/H71Y flies. We also compared MN-I2 characteristics for this allele with dsod1^null/null and dsod1^+/+ animals. Newly eclosed dsod1^H71Y/H71Y flies show normal arborization similar to dsod1^+/+ (Fig. 3B). Given the relatively short life spans of both mutant dsod1 alleles, we aged flies over a modest 10-day period. We found dsod1^H71Y/H71Y mutants underwent a retraction in axon projections, which was apparent by day 6 (Fig. 3C). Homozygous dsod1^null/null flies also showed arbor reductions in older flies at 10 days post-eclosion. In addition, the axon arbors of both genotypes appeared disorganized by day 10, lacking a stereotyped pattern with very fine axonal projections (Fig. 3B, arrowheads). In addition, there were broken branches and growth cones or synaptopodia associated with axon sprouting or de novo bouton formation as shown in Fig. 4 (Ataman et al., 2008). There was an overall age-dependent reduction in bouton number in aged flies for all genotypes; however, aged dsod1^H71Y/H71Y and dsod1^null/null had fewer boutons than aged dsod^+/+ controls (P<0.05) (Fig. 3D). Bouton sizes associated with MN-I2 are ∼1.2 µm in sod^+/+ flies and this size did not change over a 10-day period. In contrast, we found boutons became significantly enlarged or swollen in sod^H71Y/H71Y and to a lesser extent in sod^null/null mutants (Fig. 3E–G). The similar axonal phenotypes in aged-match dsod1^H71Y/H71Y and dsod1^null/null suggest bouton changes were due to loss of dsod1 function rather than H71Y associated gain-of-function. Additionally, we found similar branching and bouton sizes in aged dsod1^H71Y/+ heterozygotes to age-matched controls that further supports a loss-of-function phenotype (Fig. S3).

In Drosophila, functional glutaminergic NMJs contain post-synaptic glutamate receptors clustered by concerted action of several proteins including NETO and DISCS-LARGE (DLG) (Chen and Featherstone, 2005; Kim et al., 2012). To better characterize synapses in sod mutants, we performed ICC using anti-HRP and anti-DLG antibodies. Surprisingly, we noticed some of the HRP-staining axonal branches contained boutons that lacked surrounding DLG expression, mostly in aged sod mutants (Figs 4 and 5). Ghost boutons that lack DLG in the subsynaptic reticulum have been characterized mainly in larvae and are thought to represent non-functional synapses (Ataman et al., 2008). To quantify the extent of boutons which lacked DLG, we counted DLG positive and negative boutons and found significant (P<0.05) decreases in aged dsod1^H71Y/H71Y and dsod1^null/null mutants relative to aged-matched dsod1^+/+ as shown in Fig. 5 (53% and 51% for dsod1^H71Y/H71Y and dsod1^null/null, respectively, versus 78% for dsod1^+/+). We also observed statistically significant age-related decreases in dsod1^null/null mutants between newly eclosed and day 10 flies. Lack of associated DLG was found along entire branches and branches that showed bouton-to-bouton variability. Many of the DLG-negative branches also showed weak HRP staining and these boutons appeared to be relatively small (arrows in Figs 4 and 5B and C). Aged sod^H71Y/H71Y mutants were also enriched for boutons with filopodial extensions as shown in Fig. 5B and B″ and were also apparent in fine branches with weak HRP expression (Fig. 4, day 10). We also compared DLG coverage along axonal branches in dsod1^H71Y/+ aged flies and found strong DLG expression similar to dsod1^+/+ (Fig. S3). Thus age-dependent changes in DLG likely result from dSOD1 loss-of-function.

Flies homozygous for G85R or H48R alleles result in pharate adults that die during adult eclosion or die within 48 h after eclosion, respectively. Given these severe phenotypes, we assessed changes in MN-I2 for G85R and H48R alleles in pharate adults. Flies homozygous for either mutant dsod1 allele showed highly disorganized MN-I2 branching compared to dsod1^+/+ flies. For dsod1^G85R/G85R, there were numerous broken axons (arrow) and also branches that lacked DLG expression (Fig. S4A). Boutons that lacked DLG were quantified and trended lower for G85R, although this difference was not statistically significant (P=0.07; Fig. S4B). Finally, we found enlarged boutons relative to controls for both H48R and G85R alleles. Thus, these two most severe dsod1 alleles also produce early changes in axon morphology.

We next assessed boutons for active zones across genotypes in aged animals by monitoring bruchpilot (BRP). In sod^+/+ animals, BRP was mainly localized to boutons as expected; however, there were also parts of MN-I2 that contained BRP positive puncta but lacked distinct varicosities, including the main MN-I2 axon branch emanating from fascicle. This observation was evident across genotypes and may represent a difference in the location and morphology of NMJs between larval and adult MNs. Across genotypes, we found nearly all morphologically distinct boutons contained BRP-positive active zones; however, very large varicosities present in sod mutants showed weak or diffuse BRP staining (for example, see Fig. 6, H71Y inset). Furthermore, fine branches (arrows) that contained few boutons showed little or no BRP staining.

An accumulation of polyubiquitinated proteins as a result of decreased proteasome function and autophagy is a characteristic of ALS and other neurodegenerative diseases (reviewed in Blokhuis et al., 2013). We assessed the extent of polyubiquitinated protein aggregation in Drosophila sod mutants by performing ICC with an anti-polyubiquitin antibody (FK2) and quantifying puncta. In day 0 adults, there were relatively few FK2 positive puncta present and no statistically significant differences in numbers of puncta across genotypes (Fig. 7A,B). However, FK2 positive aggregates increased in aged sod^H71Y/H71Y versus sod^+/+ animals (P<0.05). We also found statistically significant increases in FK2-positive puncta between newly eclosed and aged sod^H71Y/H71Y (P<0.05). Furthermore, there was an increased in polyubiquitinated aggregates in aged sod^H71Y/H71Y homozygotes compared to sod^H71Y/+ heterozygotes (Fig. 7C). We found no differences in polyubiquitinated protein levels between sod^H71Y/+ heterozygotes and sod^+/+ (loxP) consistent with recessive phenotypes (data not shown). As shown in Fig. 7, puncta are localized to muscle fibers (arrows) but can also be found in distal regions of axons and in boutons (Fig. 7A′). These results indicate that there is an age-dependent accumulation of polyubiquitinated aggregates in sod^H71Y/H71Y. Such aggregates are often associated with changes in mitochondrial morphology and function; and mutant SOD1 are associated directly with mitochondrial membranes (Deng et al., 2006). We assessed age-dependent changes in mitochondrial morphology and observed swollen mitochondria across the TLM muscle in aged versus young dsod1^H71Y/H71Y mutants (Fig. 7D). Furthermore, there were also observed readily apparent changes in aged dsod1^H71Y/H71Y mutants compared to wild-type controls. Mitochondrial swelling in dsod1^nul/nulll animals was also apparent but qualitatively less severe than age-matched dsod1^H71Y/H71Y flies. Interestingly, mitochondria in sod^G85R/G85R homozygotes were similar to controls suggesting morphological changes in mitochondria may be age-related in dsod1 mutants (Fig. S4D).

This work characterizes changes in the morphology of a single MN that innervates the leg TLM in a Drosophila model of ALS. The MN-I2 arborization is less complex than many other adult MNs and possesses modest numbers of branches and boutons, making it easier to quantify changes. Furthermore, as one of several MNs that innervate the TLM, it likely plays an important role in locomotion. We saw subtle differences between MN-I2 arbor complexity in males versus females based on Sholl analysis. Other neuronal classes also show sexually dimorphic morphologies to drive sex specific behaviors (Goto et al., 2011). For leg MNs, sex-specific courtship, mating and egg-laying behaviors necessitate different patterns of innervation (Stockinger et al., 2005). However, it is unclear if or how these differences in branching patterns of MNs affect function. Our characterization showed that MN-I2 forms a stable branching pattern over a 10-day period in sod^+/+adults. While wild-type flies live for ∼90 days, we find ∼35% of dsod1^H71Y/H71Y and ∼60% of dsod1^null/null mutants are alive at 10 days, respectively (Sahin et al., 2017). From eclosion through day 10, sod mutants lose locomotor capabilities with a prominent metathoracic leg dragging phenotype. Although many factors contribute to walking behaviors in Drosophila including motor, sensory and muscle inputs, there is reasonable correlation between age-dependent deficits in locomotion for dsod1^H71Y/H71Y and reduction in arborization by day 6. Although ∼30% of dsod1^H71Y/H71Y homozygotes showed abnormal walking phenotypes at eclosion, MN-I2 arborization appeared similar to wild-type morphologically. We did not assess neuronal function by electrophysiology, which would provide an important readout, and these studies await future analysis. It should be possible to perform muscle recordings across genotypes and ages to determine the extent to which differences in motor neuron arborizations affect innervation and contractile properties of the muscle. A characterization of neurons innervating leg muscles has been published recently and found that an ordered recruitment of MN activation controls muscle contraction (Azevedo et al., 2020).

ALS is associated with age-dependent denervation. Motor loss begins at presymptomic stages and is associated with concomitant spouting from neighboring axons (Martineau et al., 2018) (Clark et al., 2016) (Hegedus et al., 2007, 2009, 2008). We find bouton loss occurred with age in dsod1 mutants and occurred along with an overall shortening of total axon length. Furthermore, we observed numerous examples of filopodia, particularly in dsod1^H71Y/H71Y suggesting there may be additional regions of axonal outgrowth. In Drosophila, such filopodia are also associated with retraction as seen when the nervous system is remodeled during metamorphosis (Liu et al., 2010). We cannot rule out this possibility given an overall reduction in axon length. During metamorphosis, DLG expression becomes diffuse and boutons become enlarged during NMJ loss. Aged dsod1 mutants share these characteristics suggesting that NMJs may be similarly dismantled. Interestingly, flies overexpressing wild-type or ALS-associated mutant FUS in MNs show >50% loss of DLG and reduced or diffuse BRP localization within boutons (Machamer et al., 2014). These results may indicate a set of shared molecular events when NMJ destruction is triggered by multiple mechanisms.

While adult MNs undergo age-dependent NMJ loss, larval MNs show only minor defects (Held et al., 2019; Sahin et al., 2017). For dsod1^G85R/G85R mutants, larvae crawl slower but possess similar number of boutons and similar evoked and spontaneous excitatory post-synaptic potentials (eEPSP and sEPSP) compared to wild-type flies. In contrast, MNs innervating the adult abdomen have fewer boutons, fewer sEPSPs, and less periplasmic DLG staining. Because dsod1^G85R/G85R mutants die prior to adult eclosion with morphological leg defects, it is unclear if adult MN development is disrupted or there is rapid NMJ loss after normal outgrowth. Larval MNs may be spared due to maternal dSOD⁺ deposition or because metamorphosis occurs before slow degenerative processes manifest. Alternatively, it is possible that there are molecular and/or physiological differences between adult and larval MN pools making the former differentially susceptible. Other studies have begun to identify stage specific MN properties (Rivlin et al., 2004).

ALS is non-cell autonomous and MN loss involves dysregulation of glia, muscle and MNs. We observed an accumulation of polyubiquitinated aggregates in muscle and distal MN processes. While the majority of ALS studies focus on protein aggregation in brain and spinal cord MNs, muscle aggregates are also detected for TDP-43 and C9ORF72 for both sporadic and familial ALS (Lynch et al., 2019; Mori et al., 2019). Furthermore, muscle-specific expression of mutant SOD1 causes MN degeneration (Dobrowolny et al., 2008; Wong and Martin, 2010). These results argue that muscle plays an important role in disease pathogenesis. Furthermore, maintenance of the NMJ requires both anterograde and retrograde signaling between muscle and MN (James et al., 2019; McCabe et al., 2003; Penney et al., 2016; Budnik and Salinas, 2011). Although not directly assessed in this study, it is possible that pathological events such as dysregulation of retrograde signaling contributes to NMJ loss in sod mutant flies. Phosphorylation of DLG through PAR-1 kinase regulates synapse formation and transmission and dlg mutant larvae have decreased bouton numbers (Zhang et al., 2007). It is therefore possible that reduction of DLG in dsod1 mutants causes our observed decreased bouton numbers in MN-I2. There are also defects in MN retrograde signaling at presymptomic stages in ALS models (Gibbs et al., 2018).

There is a strong association between SOD1 protein aggregation potential and ALS progression for the spectrum of sod1 gene mutations and a good correlation exists between aggregation potential and survival times across a broad spectrum of sod1 mutations (Lang et al., 2015). However, most sod1 mutations also decrease enzymatic activity and enzyme half-life with varying degrees (Borchelt et al., 1994; Saccon et al., 2013). Interesting, mutant SOD1^A4V retains some enzymatic activity yet causes an aggressive form of ALS. In CRISPR-edited human cells and humanized yeast strains, heterozygous SOD1^A4V/+ cells showed increased lipid peroxidation indicative of oxidative damage and substantial aggregate formation despite retaining SOD1 activity (Brasil et al., 2019). These data along with numerous other reports show aggregation and oxidative stress are linked. Thus, there is apparent oxidative damage even when SOD1 retains enzymatic activity. Mutant SOD1-H71Y protein disrupts Zn²⁺ binding critical for enzymatic activity rendering the mutant protein inactive (Cruz-Garcia et al., 2017). We have confirmed lack of enzymatic activity previously in sod^H71Y/H71Y Drosophila (Sahin et al., 2017). Thus, increased oxidative damage associated with loss of SOD1 function likely contributes directly to NMJ pathology for this and other disease-linked sod alleles where enzymatic activity is reduced. Similar studies describe a distal axonopathy in SOD1-knockout mice, which includes terminal axon swelling aberrant sprouting and phosphorylated neurofilament accumulation (Fischer et al., 2012). The fact that we observe similar NMJ phenotypes between sod^H71Y/H71Y and sod^null/null flies indicates that oxidative damage resulting from loss-of-function likely plays a significant role in NMJ pathology. However, we did find a few notable phenotypic differences between H71Y and null alleles. In sod^H71Y/H71Y flies, frequency of leg-dragging was greater, arborization decreases occurred earlier, and bouton sizes were larger relative to age-matched sod^null/null. Thus, our data indicate a combination of gain- and loss-of-function phenotypes in sod^H71Y/sod^H71Y flies. Other groups have suggested that ALS caused by sod1 mutations is due to both gain and loss-of-function and our data generally support this model (Baskoylu et al., 2018; Saccon et al., 2013; Sau et al., 2007).

Here we use a novel leg dissection preparation to characterize the axon arbor from a single MN in Drosophila. The MN-I2 motor neuron is easily identifiable and tractable to morphometric and molecular analysis making further studies of aging and slow degenerative processes possible. In dsod1 mutant adults, we show that MN-I2 undergoes an age-dependent reduction in axon length accompanied by a loss of boutons. Boutons remaining in MN-I2 are enlarged and periplasmic DLG is reduced or absent. Finally, we find an accumulation of polyubiquitinated aggregates in aged flies. Although overt MN loss does not occur in mutant sod flies, these observations are consistent with denervation as seen in other ALS model systems and confirms the robustness of our model in recapitulating important cellular aspects of human ALS.

Drosophila sod alleles were described previously (Sahin et al., 2017). Fly stocks were reared on standard sucrose, cornmeal, agar media at 23°C with a 12:12 light:dark cycle. For aging experiments, newly eclosed flies were collected daily and changed to new media every 1–3 days.

Adult walking assays were performed as previously described (Diaper et al., 2013). Images were captured with a compound microscope (100x) and brightness and contrast were adjusted as needed.

Metathoracic legs were cut at the coxal segment and fixed in 4% formaldehyde for 45 min prior to dissection. After washing 3× in PBS, a cuticle patch along the femur was removed exposing underlying tissue and samples were post-fixed for 15 min in 4% formaldehyde. Antibody staining protocols followed standard methods and samples were mounted in slow-fade diamond mount (Molecular Probes) (Patel, 1994). Antibodies and stains used were anti-horse-radish peroxidase (#115-035-166, Jackson ImmunoResearch, 1:550), anti-dlg (4F3, Developmental Studies Hybridoma Bank, 1:200), phalloidin-633 (Molecular Probes, 1:1000), mouse anti-FK2 monoclonal (#BML-PW0150-0025, Enzo Life Sciences, 1:200); mouse anti-Bruchpilot monoclonal (nc82, Developmental Studies Hybridoma Bank, 1:20), mouse anti-ATP5a (#ab14748, Abcam, 1:200). Imaging was performed by confocal microscopy (Olympus FV1000) capturing z-stacks for all samples. Image processing occurred in ImageJ (FIJI) and consisted of cropping images for areas of interest and altering brightness/contrast as needed for publication. Where necessary image size was increased using linear transformation. Image stitching was performed in ImageJ as needed (Preibisch et al., 2009).

Sholl analysis was performed on MN-I2 to assess differences in motor neuron structure as described previously (Ferreira et al., 2014; Sholl, 1953). Briefly, the center was specified using the straight-line method starting at the branch point where MN-I2 exits the main axonal bundle and the ends at the most distal part of the motor neuron. The average number of intersections was recorded for left and right metathoracic legs for male and female recently eclosed flies. Student's t-test was performed between left/right and male/female samples.

Measurement of axon lengths were performed with the Simple Neurite Tracer plugin for ImageJ (Longair et al., 2011). Bouton counts and sizes were also determined within ImageJ using the count and measure functions, respectively. Data were evaluated by Levene's test for homogeneity of variance and Shapiro's test for assessing normality. Bouton sizes were transformed using natural-log transformation to normalize the distribution of bouton sizes. For bouton counts, sizes, and axon lengths a two-way ANOVA with a post hoc Tukey test was performed with respect to age and genotype to account for multiple comparisons.

Western blotting was performed as previously described (Sahin et al., 2017). Briefly, metathoracic legs were isolated and homogenized at 4°C in 40% RIPA, 60% lammeli+5% β-mercaptoethanol. The samples were spun for 10 min at 16,000×g and the supernatant was transferred to a new microcentrifuge tube. The supernatant was then boiled for 10 min, cooled on ice and then loaded onto 4–20% SDS-PAGE gels. Following electrophoresis, protein was transferred to blotting membrane, and blocked for 2 h at room temperature in superblock (Thermo Fisher Scientific) or 1% BSA in tris-buffered saline +0.01% Tween (TBST). Following primary antibody incubation and wash, blots were incubated with anti-peroxidase secondary antibodies (Jackson ImmunoResearch,1:10,000) for 2 h and washed three times in TBST. Blots were developed with SuperSignal West Pico Plus Chemiluminescence (Thermo Fisher Scientific) and imaged using a Chemidoc digital imaging system (Bio-Rad).

Polyubiquitin-positive puncta detected by ICC were quantified across genotypes and ages using the Analyze Particle plugin within ImageJ on 1400 µm areas encompassing MN-I2 using images captured by confocal microscopy at the same magnification (60× objective with 1.4× zoom).

We would like to thank the Biology department at Rhode Island College, and particularly Ron Pitt for support and Eric Roberts for assistance with confocal microscopy. Monoclonal antibodies anti-dlg and anti-brp were developed by UC-Berkeley and Universitaetsklinikim Wuerzburg, respectively, and were obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, USA. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study.