Enhancing GABAergic Transmission Improves Locomotion in a Caenorhabditis elegans Model of Spinal Muscular Atrophy

Abstract Spinal muscular atrophy (SMA) is a neuromuscular disease characterized by degeneration of spinal motor neurons resulting in variable degrees of muscular wasting and weakness. It is caused by a loss-of-function mutation in the survival motor neuron (SMN1) gene. Caenorhabditis elegans mutants lacking SMN recapitulate several aspects of the disease including impaired movement and shorted life span. We examined whether genes previously implicated in life span extension conferred benefits to C. elegans lacking SMN. We find that reducing daf-2/insulin receptor signaling activity promotes survival and improves locomotor behavior in this C. elegans model of SMA. The locomotor dysfunction in C. elegans lacking SMN correlated with structural and functional abnormalities in GABAergic neuromuscular junctions (NMJs). Moreover, we demonstrated that reduction in daf-2 signaling reversed these abnormalities. Remarkably, enhancing GABAergic neurotransmission alone was able to correct the locomotor dysfunction. Our work indicated that an imbalance of excitatory/inhibitory activity within motor circuits and underlies motor system dysfunction in this SMA model. Interventions aimed at restoring the balance of excitatory/inhibitory activity in motor circuits could be of benefit to individuals with SMA.


Introduction
Spinal muscular atrophy (SMA) is an autosomal recessive disease in infants and children, which is characterized by progressive degeneration of the motor neurons resulting from a reduction in abundance of functional survival motor neuron (SMN) protein (Lefebvre et al., 1997). SMN cycles between the nucleus and cytoplasm, and plays a critical role in assembly of the ribonucleoproteins (RNPs) required for pre-mRNA splicing (Fischer et al., 1997). In neurons, SMN is also found in axons and is thought to be important for RNA transport (Rossoll et al., 2003). Although SMN is ubiquitously expressed in all cells, clinically, the neuromuscular system displays the most prominent pathology in SMA. The cell type-specific pathology of disease is one of the more enigmatic aspects of SMA, it may reflect relative differences in tissue requirements for SMN-mediated biological processes Zhang et al., 2008).
A variety of SMA animal models have been established including mice, zebrafish, flies, and worms. These platforms have provided insight into a number of critical pathophysiological questions (McWhorter et al., 2003;Oprea et al., 2008;Hua et al., 2011;Imlach et al., 2012). First, in which cell or cell types is SMN required for health and maintenance of the neuromuscular system? In mice, cell type-specific knock-out (KO) or cell type-specific rescue in an SMN deficient background suggests that the SMA phenotype is a result of loss of SMN from motor neurons Zhou et al., 2016), other neurons (Paez-Colasante et al., 2013), and muscle cells (Cifuentes-Diaz et al., 2001). In addition, peripheral tissues could be critical players, too (Hua et al., 2011). Second, is SMN required continuously to maintain neuromuscular system integrity or is there a critical window? Phenotypes of the conditional KO mice indicate that the period of neuromuscular junction (NMJ) formation/maturation (especially between postnatal day 10 and postnatal day 20) has a particularly high requirement for SMN, and adult animals have a substantially less, but non-trivial, requirement for SMN (Swoboda, 2011;Kariya et al., 2014). Third, which biological activities subserved by SMN are necessary for proper functioning of neuron? This has been challenging to unravel, but defects in splicing specificity are thought to be the main disease drivers (Zhang et al., 2008).
Humans have two copies of the SMN gene (SMN1 and SMN2) that differ in a C/T nucleotide inside exon 7, which alters its splicing patterns. This critical difference results in an inclusion of exon 7 in the majority of SMN1-derived mRNA but an exclusion of exon 7 in the most of SMN2derived mRNA. SMN protein lacking sequences encoded by exon 7 is un stable (Burnett et al., 2009). SMA is caused by the low levels of SMN protein, which results from loss of SMN1 and variable compensation by the unstable products of the SMN2 locus (Monani et al., 2000;Le et al., 2005). These observations on the molecular pathophysiology of SMN have therapeutic implications. In a mouse model of SMA, application of antisense oligonucleotide (ASO) therapy aimed at increasing the expression of exon 7-containing SMN from the SMN2 locus have a major impact on disease (Hua et al., 2011), and human trials using this approach are promising (Finkel et al., 2016;Mercuri et al., 2018). Nonetheless, this therapeutic approach is problematic because ASOs do not cross the blood brain barrier, and intrathecal administration of ASO to fragile patients is a clinical issue. Non-SMN modifiers of the SMA phenotype have been identified (i.e., plastin 3 and ROCK). Targeting these molecules and the cell biological processes they impact has therapeutic potential (Oprea et al., 2008).
In an effort to identify new disease modifiers we asked whether genes previously implicated in life span extension of Caenorhabditis elegans (Kenyon, 2010) would influence the lifespan and locomotor phenotype of smn-1 null animals. We discover reduction of the insulin/insulin-IGF signaling (IIS) pathway significantly extends lifespan and improves locomotor phenotypes in this SMA model. These effects are intriguing because reducing other genes previously implicated in lifespan extension did not extend lifespan in smn-1 null animals. Furthermore, we further demonstrate that by loss of daf-2, locomotion is improved due to enhancement of GABAergic transmission. Our data suggest that impaired GABAergic transmission is disrupted normal homeostatic locomotion in SMA disease model. Interventions aims at restoring the balance of excitatory/inhibitor activity in motor circuits could be benefits to individuals with SMA.

RNAi knock-down and survival assay in C. elegans
Each RNAi (RNA Interference) colony was grown overnight in Luria broth (LB) containing ampicillin (50 g/ml), and 200 l was seeded onto NGM plates containing isopropylthiogalactoside (IPTG; 1 mM) to induce dsRNA expression at room temperature. All RNAi clones used in this study were generated as described previously in the Ahringer library (Fraser et al., 2000) and were confirmed by DNA sequencing. Longevity-promoting RNAi candidates were validated in wild-type N2 animals. For lifespan experiments, ten young adult animals were placed on RNAi plates, eggs were laid for 4 -6 h before hermaphrodites were removed. The progeny was allowed to grow on the RNAi plate until young adults (first generation). This cycle was repeated once more to obtain the thirdgeneration progeny for the lifespan assay. Survivors were counted every day. An empty vector (EV) for RNAi constructs served as a control.

Confocal images and synapse marker (puncta) analysis
Animals at L2 or Late (L2 ϩ 3 d) stage were mounted on dried 2% agarose pads with M9 buffer. A stack of fluorescent images was captured from same posterior region of the dorsal nerve cord using Olympus laser scanning confocal microscopy with 100ϫ objective lens. Stacked images were projected into a single plane using Fluoview software. The acquired images were processed and analyzed by an in house-developed PunctaAnalyzer program and R script (Hung et al., 2007).

Pyridostigmine bromide and levamisole sensitivity assay
For analysis of sensitivity to the drugs pyridostigmine bromide and levamisole, three replicates of at least 20 worms per genotype were placed on NGM plates supplemented with either 100 mM pyridostigmine bromide (Sigma) or 100 M levamisole (Sigma). Number of worms paralyzed on each plate was counted every 30 min to a total 360-min duration. Paralyzed worms were identified as those failing to response after 5 s of stimulation by plate-tapping and tail-prodding. The percentage paralyzed was averaged at each time point and plotted graphically using Log-rank (Mantel-Cox test) in PRISM.

Quantitative RT-PCR analysis
We hand-picked 500 worms of the various genotypes and at the examined ages for extraction of total RNA using TRIzol (Invitrogen) and RNeasy Mini kit (QIAGEN). A total 1 g of DNase-treated RNA were reverse transcribed using iScript Reverse Transcription Supermix (Bio-Rad) according to the manufacture's protocol. The cDNAs were subjected to real-time qPCR in a total volume of 25 l, containing 1ϫ Power SYBR Green PCR Master mix (Applied Biosystems) and 200 nM primer. The qPCRs were amplified and analyzed in triplicate using StepOne Real-Time PCR system (Invitrogen). Relative mRNA level of the genes of interest were normalized against ␣ tubulin (tba-1) as an endogenous control. Primer sequences for amplification of target mRNA were as following: smn-1, 5=-ATACCTCGATGCCATTTCCA-3= and 5=-ATCCGCT-CATGTACCAGCTC-3=; tba-1, 5=-AAGATGCCGCCA ACAACTAC-3= and 5=-CCTCCTCCGAATGAATGAAA-3=.

Swimming assay and biomechanical profile analysis
We captured movies of swimming nematodes in sealed fluidic chambers that are 2 cm in diameter and 1 mm in depth. Imaging is performed using bright-field microscopy and a high-resolution CCD camera at 30 frames per second. We discard all movies that exhibit nematode-wall interactions and interrupted or irregular beating, as well as movies with fewer than four (worm) beating cycles. Excluding discarded movies, we obtain an average of 20 movies of unique individuals per group tested, with a minimum of 11.
Movies of swimming nematodes are analyzed using in-house software with slight modification from the published software which was previously used and validated for a study on biomechanical profiling of C. elegans genotypes (Krajacic et al., 2012). The original software is freely available as supplemental material from the Genetics journal website (http://www.genetics.org/content/191/ 3/1015.supplemental). The modified version of software in this study is freely available via the University of Pennsylvania Complex Fluids Lab website (http://arratia.seas.upenn.edu/software.html). The modifications allow the code to be used for the study of immature rather than adult C. elegans. We have implemented changes that allow us to accurately measure the biomechanical properties of younger and therefore smaller and thinner worms by replacing parameters valid for adult, approximately millimeter-length C. elegans with parameters calculated directly from measured lengths and diameters.
The software uses image processing techniques to track the swimming motion of C. elegans by extracting the body-shape of the swimmer and then computing the nematode's biomechanical properties, including frequency, amplitude, speed, force, and power (Krajacic et al., 2012). In brief, the nematode's body is separated from the background of the image, and its centroid and body shape in each frame in the plane of motion is computed as well as the locations of its head and tail.   RK122 smn-1(rt248)/hT2;daf-2(e1370) RK123 smn-1(ok355)/hT2; snf-11(ok156) Differentiating the centroid position in time yields the forward swimming speed U. Next, we divide the centerlines into segments, which are tracked in space and time to create a parameterized body shape s͑x, y, t͒, where x and y define the plane of motion and t represents time. The nematode's bending curvature is defined as the change in local orientation along the body of the worm. When plotting curvature as a function of time, diagonal stripes of alternating positive and negative curvature clearly show the periodicity of the nematode's swimming gait, where the beating frequency is defined as f ϭ 1/T, where T represents the time to complete one beating cycle.
We also calculate the normalized curvature as a function of position along the nematode by first averaging the local curvature in time to obtain Ϫ͑s͒ for each individual nematode. We then take the magnitude, or absolute value, of Ϫ͑s͒, the average across each population, and normalize by the average curvature at the head of the worm to compare across genotypes. Note that we neglect the first and last 5% of the nematode's parameterized body shape due to its decreasing diameter and increasing transparency which introduce uncertainties in our curvature measurements.
Moreover, we use the measured body shapes and velocities to compute factors that might differentiate the behavior of nematodes with SMA from healthy wild-type organisms. We use a simple hydrodynamic model, resistive force theory (Gray and Hancock, 1955), to estimate the propulsive force of the nematode (Sznitman et al., 2010). Since C. elegans are self-propelled and force-free (Lauga and Powers, 2009), the sum of its propulsive and drag forces must be zero such that where F P and F D are the propulsive and drag forces acting on the nematode, respectively. We can then estimate F P by computing F D from RFT; for each body segment, we can decompose its force contribution as: where F N and F T are the normal and tangential components of the drag force, C N and C T are the normal and tangential components of the drag coefficient, and u N and u T are the normal and tangential components of the velocity of each body segment, respectively. The drag coefficients C N and C T are given by Lighthill's refinement of the original RFT coefficients (Gray and Hancock, 1955;Lighthill, 1976). Performing the integral over the surface of the worm, we can estimate propulsive force (Lauga and Powers, 2009;Sznitman et al., 2010) using: The calculation of period-averaged propulsive force provides an estimate of muscle (mechanical) output, and more intuitively, as an indirect measure of strength.
Furthermore, we can compute the power, or rate of energy expenditure, of C. elegans by multiplying the component of the propulsive (or drag) forces for each segment by its local velocity components: A period-averaged calculation of power is therefore an estimate for how much mechanical energy is consumed by C. elegans for locomotion.

Statistics
Statistical analysis was performed using GraphPad Prism. For data containing more than three groups of non-normally distributed data, a non-parametric Kruskal-Wallis test with Dunnett's multiple comparisons test was used. For normally distributed data (puncta analysis), oneway ANOVA with the post hoc Tukey HSD was applied for multiple comparisons. For survival analysis, a Log-rank (Mantel-Cox) test was used. The threshold for significance was set to p Ͻ 0.05. For normalized body curvature, multiple t tests followed by Holm-Sidak method correction were used.

Inactive daf-2/IGF-1-like signaling pathway extend lifespan in SMA worm models
To identify new SMA disease modifier, we studied a C. elegans smn-1 null mutant ok355 (gene name C41G7.1). The null allele is maintained in a heterozygous state using the hT2g balancer chromosome and heterozygous animals (smn-1(ok355)/hT2g) are phenotypically normal. Maternally contributed SMN message to the developing embryo allows for the smn-1(ok355) mutant animals to develop and mature until their premature demise before adulthood. Animals homozygous for the null allele smn-1(ok355) have two prominent phenotypes: (1) impaired locomotion (e.g., uncoordinated or "unc"), and (2) death at 7-8 d of life at 20°C (Briese et al., 2009). Here, we focused on these phenotypes due to their potential relevance to the SMA disease.
To address these possibilities, we conducted real-time quantitative PCR to monitor smn-1 RNA level in wild-type N2, smn-1(ok355), and smn-1(ok355);daf-2(e1370) mutant animals at various developmental time points (Fig.  1C). In wild-type N2 animals, we found a progressive decline in the SMN messenger RNA abundance from day 2 to day 7 on hatching. Compared to smn-1 RNA level from N2 animals, barely detectable smn-1 RNA levels were found in both smn-1(ok355) and smn-1(ok355);daf-2(e1370) strains, across multiple developmental time points. This result suggests that loss of daf-2 activity prolongs lifespan in smn-1(ok355);daf-2(e1370) strain despite the lack of smn-1 mRNA. The absence of good tools for monitoring SMN protein levels leaves a caveat that loss of daf-2 might slow SMN protein degradation. In mammalian cells, the half-life of full length SMN protein is ϳ4.5 h (Burnett et al., 2009). It seems unlikely that animals lacking SMN mRNA for 5ϩ d after hatching would still have functional SMN protein. Thus, it suggests that loss of daf-2 renders smn-1(ok355) mutant animals insensitive to the absence of SMN protein.
Taken together, our results reveal that not every longevity promoting pathways identified in C. elegans are capable to provide survival benefit in smn-1 null animals. Only reduction of daf-2 signaling pathway has such unique ability to extend the life of animals lacking SMN-1. The beneficial effects of reducing daf-2 appear to be independent of changes in SMN mRNA and probably protein.

Reduced daf-2 signaling improves the locomotor activity of smn-1 mutant animals
Based on these observations with daf-2 LOF we explored our main interest -the impaired locomotion of the smn-1(ok355) mutant animals. (Briese et al., 2009). Although smn-1(ok355) mutant animals display impaired locomotor activity, there is neither neuronal death nor gross anatomic abnormalities in motor circuit wiring (Briese et al., 2009). This suggests that functional deficits within the circuitry precede anatomic abnormalities, an observation commonly made in neurodegenerative disease models (Ilieva et al., 2008;Shakkottai et al., 2011). An alternative explanation is that relevant, but subtle anatomic defects require higher resolution examination. In the mouse SMA models, the circuitry devoted to controlling motor neuron activity is perturbed before the detection of NMJ abnormalities, and that these circuit abnormalities correlate with impairments in motor behaviors (Mentis et al., 2011). Data are lacking for quantifying the effect of loss of C. elegans smn-1 on locomotion or NMJ integrity and the effect of modifier genes on these phenotypes.
We addressed these questions using automated image tracking and processing coupled with hydrodynamic models to extract quantitative kinematic (e.g., swimming speed) and dynamic (e.g., propulsive force) properties of C. elegans during swimming in a balanced salt solution (M9; Brenner, 1974). Previously, smn-1(ok355) mutant animals have been shown to exhibit a progressive decline in a single measure -thrashing rate after the second larva (L2) stage (Briese et al., 2009). With this in mind, we monitored the swimming behavior of each experimental group at two different stages, L2 and "late" (L2 ϩ 3 d), and determined if a reduction of daf-2 signaling rescues locomotion deficit in smn-1(ok355) mutant animals. Choosing the best comparison groups for the late stage is not straightforward; late smn-1(ok355) mutant animals cannot be compared with 5 days-old N2 animals because they have arrested at an earlier larval stage. We selected control groups based on the most biomechanically relevant developmental parameter, the body length. The typical body lengths for the two experimental groups are 0.39 Ϯ 0.01 mm (L2) and 0.63 Ϯ 0.08 mm (late; Table 2).
Simple observation of the swimming gait of individuals from each genotype over one second demonstrates that smn-1(ok355) mutant animals move slowly and beat less frequently than N2 and daf-2(e1370) mutant animals (Fig.  3A). In the daf-2 mutant background, smn-1(ok355) mutant animals demonstrate a qualitative improvement in swimming gait, leading to a pattern that more closely resemble N2 animals. Next, we present kymographs of the worm's body curvature over one second of forward swimming ( Fig. 3B; for more detail, see Materials and Methods). The stripes show body curvature progressing down the length of the animal from head to tail; the slope of these stripes represent the wave speed, or the rate at which waves of curvature progress along the body of the worm. The rate at which the pattern repeats itself repre-Figure 2. Enhanced survival benefits in SMA worm models are mediated by the daf-2/daf-16 pathway. A, Knock-down daf-16 by RNAi feeding shortens lifespan of smn-1(ok355) mutant animals by 12.5% when compared to EV (p Ͻ 0.001, Log-rank test). B, smn-1(ok355);daf-2(e1370) mutant animals live three more days than smn-1(ok355) mutant animals, and this beneficial effect is suppressed by feeding RNAi targeting daf-16 (p Ͻ 0.001, Log-rank test). C, smn-1(rt248);daf-2(e1370) mutant animals live 15 d, and this beneficial effect is suppressed by feeding RNAi that targets all isoforms of daf-16 (p Ͻ 0.001, Log-rank test). Feeding RNAi that targets daf16a or daf16f isoforms also suppress the longevity of smn-1(rt248);daf-2(e1370) mutant animals although the effect is less robust than the pan daf-16 RNAi effect (p Ͻ 0.001, Log-rank test). D, Overexpression of daf-16 using TJ356 (daf-16::gfp) strain prolongs lifespan of smn-1(ok355) mutant animals by 43.8% when compared to smn-1(ok355) mutant animals (p Ͻ 0.001, Log-rank test).
sents the frequency f of the worms beating. When examining the curvature kymographs of each genotype, we see that smn-1(ok355) mutant animals have a region along their body near the tail that exhibits no curvature. This region stretches from approximately the middle of the worm to just before the tail (0.5 Ͻ s / L Ͻ 0.8). This suggests a LOF of neuromuscular function in that region, since repeated contraction and relaxation would bend the worm locally, therefore producing curvature. The "failure of curvature" phenotype is reversed in the daf-2 mutant background (i.e., animals with the smn-1(ok355);daf-2(e1370) genotype).
To interrogate this failure of curvature issue more closely, we recall that the body curvature of the nematode is dependent on its local orientation and its parameterized body shape. By quantifying the body curvature of all the nematodes for each genotype, we systematically probe this aspect of worm locomotion with high sensitivity. The curvature of a worm appears to have nodes when we take an average in time over each population ( Fig. 3C; Extended Data Fig. 3-1). For the N2 and daf-2(e1370) mutant animals, we observed the mean normalized curvature decreases from head to tail. The daf-2 group shows a significant lower curvature at its tail region compared to the N2 group. The curvature pattern in the smn-1(ok355) mutant animals are very distinct from N2 and daf-2. It decreases rapidly after the worm's mid-point and then sharply increase right near the tail ( Fig. 3C; adjusted p Ͻ 0.00001 by multiple t tests and Holm-Sidak method correction). This pattern corresponds to the region of low curvature observed in Figure 3B kymographs. Moreover, we observed that smn-1(ok355);daf-2(e1370) mutant animals displayed a normalized mean curvature between those of N2 and daf-2(e1370) mutant animals. It is clear that the defects in body curvature in the midsection and tail incurred by loss of smn-1 are rescued in the daf-2 mutant background.
Next, we undertook a quantitative analysis of worm swimming using in-house biomechanical profiling algorithms first published in Genetics (Krajacic et al., 2012), and examined 13.5 h of videos at 30 frames per second (approximately yielding 570 gigabytes of data). Using prespecified criteria to exclude substandard data (as described in Materials and Methods), we ultimately processed 0.76 h of videos (corresponding to 31.9 gigabytes of data). Figure 4A shows that the body length of each group is significantly increased from L2 to late stage. There is no significant difference in body length across all groups at each stage (L2 or late stage; Fig. 4A; Table 2). The body length is the most biomechanically relevant developmental parameter and serves as the control here. Then, we begin evaluated various biomechanical properties by swimming assay. First, we measured each experimental group's swimming speed U and beating frequency f. At the L2 stage, the swimming speed of N2 animals is ϳ0.16 mm/s (Fig. 4B), and this is no different from daf-2(e1370), smn-1(ok355), and smn-1(ok355);daf-2(e1370) mutant animals (p Ͼ 0.05, Kruskal-Wallis test). With developmental progression (i.e., the late stage), wildtype N2 and daf-2(e1370) animals significantly increase their swimming speed by 63% and 65%, respectively. In contrast, at the late stage smn-1(ok355) and smn-1(ok355);daf-2(e1370) mutant animals trend toward a decrease in swimming speed, there was no statistically significant difference between swimming speed at the early and late time points. This suggests that the loss of smn-1 impairs the normal, age-dependent increase in locomotion rate.
Next, we investigated beating frequency in these animals. Although N2 animals increases modestly over time (1.9 Hz at the L2, and 2.3 Hz at the late stage, p Ͼ 0.05, Kruskal-Wallis test), both N2 and daf-2(e1370) animals did not show significant increase in frequency (Fig. 4C). The beating frequency of smn-1(ok355) mutant animals at the L2 stage is significantly higher than N2-L2 animals (2.5 vs 1.9 Hz, p Ͻ 0.05, Kruskal-Wallis test) and decreases by 32% at the late stage (e.g., 2.3-1.7 Hz, p Ͻ 0.01, Kruskal-Wallis test). We also found that the beating frequency of smn-1(ok355);daf-2(e1370) mutant animals at the L2 stage is similar to the age-matched smn-1(ok355) (2.6 vs 2.5 Hz). Interestingly, unlike the smn-1(ok355) group, smn-1(ok355);daf-2(e1370) mutant animals are able to maintain beating frequency at the late stage in a fashion similar to wild-type N2. A pathologic decrement in neuromuscular activity over time in the smn-1(ok355) mutant animals could account for the observed decrease in beat frequency. Such pathologic decrement does not occur in the smn-1(ok355);daf-2(e1370) mutant animals.
The analysis of these simple swimming kinematics reveals a pathologic beating frequency in the smn-1(ok355) mutant animals, this does not occur in the smn-1(ok355); daf-2(e1370) mutant animals (e.g., the swimming speed over time is unmodified). While swimming frequency and speed are important factors in describing locomotion, there are of other determinants that contribute to propulsive force and mechanical power. We used the hydrody-  -2(e1370), smn-1(ok355), and smn-1(ok355);daf-2(e1370)  namic model resistive force theory (Gray and Hancock, 1955) to estimate the propulsive force and mechanical power of C. elegans, measurements that more directly probe muscle output and the energy expended during locomotion.
We expect that as a worm develops, the propulsive force of the organism will increase. Indeed, we found that both N2 and daf-2(e1370) animals exhibited an increase in propulsive force of 39% and 83%, respectively, when they progress from L2 to late stage (p Ͻ 0.01, Kruskal-Wallis test; Fig. 4D). In contrast, for smn-1(ok355) mutant animals, the propulsive force remains the same during this developmental period, suggesting that motor output is impaired. In contrast, smn-1(ok355);daf-2(e1370) mutant animals show an increase in propulsive force by 24% (p Ͻ 0.01, Kruskal-Wallis test) over this period. Thus, the normal increase in motor output as worms mature fails to occur in smn-1 mutant animals. Reduced daf-2 signaling can rescue this defect.
Next, we investigate the mechanical power expended by each group during swimming. We observed that N2 and daf-2(e1370) animals significantly increase their me- Figure 3. Reduced daf-2 signaling rescues the body curvature of smn-1(ok355) animals during swimming. A, A trace of the head and tail trajectories for typical individuals from each genotype (e.g., N2, smn-1(ok355), daf-2(e1370) and smn-1(ok355);daf-2(e1370)). There is no qualitative difference between N2 versus the daf-2(e1370) mutant animals while the smn-1(ok355) mutant animals swim more slowly and beat less frequently than N2 worms. The smn-1(ok355);daf-2(e1370) mutant animals closely resemble N2 worms. B, Curvature kymographs of representative individuals of each genotype reveals a common pattern shared by N2, daf-2(e1370), and smn-1(ok355);daf-2(e1370) mutant animals. In contrast, smn-1(ok355) mutant animals exhibits relatively little curvature (white color) compared to other genotypes at a region from approximately the middle of the body to just before the tail, implicating a regional loss of neuromuscular function. CW, clockwise; CCW, counterclockwise. C, Quantification of regional body curvature in smn-1 mutant animals. We quantified the body curvature of animals with four genotypes (N2, smn-1(ok355), daf-2(e1370), and smn-1(ok355);daf-2(e1370)) and find that the curvature of each group of worms appears to have local maxima when averaged over time (for details, see Extended Data Fig. 3-1). For the N2 and daf-2(e1370) mutant animals, the mean normalized curvature decreases from head to tail. The daf-2(e1370) group shows a significant low curvature at the tail region compared to the N2 group. The smn-1(ok355) group has a distinct curvature pattern. Its curvature decreases rapidly after the worm's mid-point and then increases sharply near the tail. This pattern corresponds to the region of low curvature observed in Figure 3B. In the daf-2 mutant background (smn-1(ok355);daf-2(e1370)), body curvature during swimming is normalized and appears similar to N2 and daf-2(e1370) mutant animals. This is quantitative evidence that the daf-2 mutant background rescues the neuromuscular defect seen in the midsection and tail of the smn-1(ok355) mutant animals. ns, no significance. chanical power (250% and 200%, respectively) as they develop (Fig. 4E). In contrast, smn-1(ok355) mutant animals produced the same mechanical power at the L2 and the late stage; in addition, the mechanical power of the smn-1(ok355) group is significantly lower than N2, late stage (0.9 vs 1.7 pW, p Ͻ 0.001, Kruskal-Wallis test). Reduction of daf-2 signaling (smn-1(ok355);daf-2(e1370)) significantly improves mechanical power when compared to smn-1(ok355) mutant animals. Mechanical power for locomotion of smn-1(ok355);daf-2(e1370) mutant animals is statistically indistinguishable from that of the late stage wild-type group (p Ͼ 0.05, Kruskal-Wallis test; Fig. 4E).
In summary, the kinematic and dynamic biomechanical properties (i.e., frequency, propulsive force, and mechanical power) show that a reduction of daf-2 signaling improves several aspects of locomotor capability in smn-1(ok355) mutant animals. While these modifications do not directly increase the swimming speed of the organism, they improved the beating patterns, muscle output, and the energy expended during locomotion.
At the L2 stage, a qualitative comparison of presynaptic marker revealed essentially no differences in puncta lo- , and (E) mechanical power (pW) for the indicated genotypes at L2 or late time points. Data shown are mean Ϯ SD, Ն10 animals per genotype were recorded. Nonparametric Kruskal-Wallis test followed by Dunn's multiple comparison test for significant; ‫ء‬ p Ͻ 0.05, ‫ءء‬ p Ͻ 0.01, ‫‪p‬ءءء‬ Ͻ 0.005, ‫‪p‬ءءءء‬ Ͻ 0.001, ns, no significance. A, Genotypes are well matched for length at the two time points. B, Age-dependent increase in speed was shown in both N2 and daf-2(e1370) mutant animals but not in smn-1(ok355) or smn-1(ok355);daf-2(e1370) mutant animals. C, Beat frequency diminishes over time in the smn-1(ok355) mutant animals and this is suppressed in the smn-1(ok355);daf-2(e1370) mutant animals. D, Age-dependent increase in propulsive force is seen in N2 and daf-2(e1370) mutant animals; this is not seen in the smn-1(ok355) mutant animals, but is rescued in the smn-1(ok355);daf-2(e1370) mutant animals. E, Age-dependent increase in mechanical power force is seen in N2 and daf-2(e1370) mutant animals; this is not seen in the smn-1(ok355) animals but is rescued in the smn-1(ok355);daf-2(e1370) mutant animals. calization and distribution among the four experimental groups (Fig. 5, top left). The quantitative analysis confirms this with the only exception that puncta intensity in L2 stage of smn-1(ok355);daf-2(e1370) mutant animals was 33% stronger than smn-1 mutant animals (p Ͻ 0.05; Fig.  6A,C,E,G). A qualitative comparison of postsynaptic   -2 activity. A, B, Quantification of puncta width in presynapse and postsynapse at L2 or late stage in different genotypes of worm. C, D, Quantification of puncta intensity in presynapse and postsynapse at L2 or late stage in different genotypes marker at L2 stage reveals a few small group differences in puncta localization and distribution among the four groups (Fig. 5, bottom left). Quantitatively, the most pronounced difference was that smn-1(ok355) mutant animals had 30% fewer puncta (Fig. 6F) and concomitantly larger gaps between these puncta when compared with N2 group (Fig. 6H). Reduced daf-2 signaling did not rescue this phenotype. Overall, at the L2 stage, all groups exhibit similar presynaptic and postsynaptic puncta patterns although a few, relatively subtle, group differences are evident.
Together, these observations reveal that young smn-1(ok355) mutant animals appear to have a normal GABAergic NMJs. Over time there are two major changes to these synapses: (1) presynaptic elements decrease in number but are restructured, and (2) the number of postsynaptic elements increase without an effect on width or GABAergic receptor content. The main effect of daf-2 reduction is to decrease postsynaptic element number to the N2 level and to increase the synaptic content of GABA receptors. It is difficult to predict the functional consequence of these diametrically opposing effects of the daf-2 mutant animals on postsynaptic GABA receptors.

Postsynaptic defects in GABAergic NMJs significantly contribute to smn-1(ok355)'s hypersensitivity to pyridostigmine bromine
To investigate the functional consequence of morphologic abnormality of GABAergic NMJs observed above, we used pharmacological tools: pyridostigmine bromide (Sleigh et al., 2011), an acetylcholinesterase inhibitor, and levamisole, a cholinergic agonist. The time course of hypercontractive paralysis of animals bathed in these agents can report the balance between muscle excitatory cholinergic and inhibitory GABAergic transmission (Mahoney et al., 2006;Dabbish and Raizen, 2011). We examined animal sensitivity to levamisole or pyridostigmine bromide at two time points: L2 stage and late stage (Fig. 6). Where applicable, we classified animals as resistant to inhibitors of cholinesterase (Ric) or hypersensitive to inhibitors of cholinesterase (Hic).
GABAergic neurotransmission (or a combination of both) would be compatible with these observations. That the HIC phenotype is suppressed in the daf-2(e1370) background suggests that equipoise in the excitation/inhibition balance has been normalized.
To the extent that changes in GABAergic transmission are contributing to these observations, we interpret our morphologic observations to indicate that: (1) reduced GABAergic transmission in smn-1(ok355) mutant animals is predominantly due to reduced number of presynaptic elements, and (2) benefit of reduced daf-2 in the smn-1(ok355) mutant animals is due to increased density of GABA receptors at synapses.

Enhancing GABAergic transmission improves locomotion of smn-1(ok355) mutant animals
If the locomotor benefit conferred on smn-1(ok355) mutant animals by loss of daf-2 is mainly due to increased GABAergic transmission, a functional enhancement of the GABAergic system in the smn-1(ok355) mutant animals should phenocopy loss of daf-2 function. To test this idea, we identified two strains of worms that are likely to exhibit enhanced GABAergic neurotransmission without affecting the baseline swimming behaviors: RM2710 (snf-11(ok156)) and FY297 (oxls22(UNC-49::GFP)). The snf-11(ok156) mu-tant animals lack the sodium-dependent GABA transporter (which clears synaptic GABA). These animals are aldicarbresistant supporting the view that they have enhancement of GABAergic transmission (Mullen et al., 2006). The overexpression of multiple subunits of the GABA receptors in oxls22(UNC-49::GFP) animals probably have enhanced GABAergic transmission, although to our knowledge, this has never been formally tested.
Next, we analyzed the curvature in the second set of animals. The four experimental groups were: N2, smn-1(ok355), snf-11(ok156), and smn-1(ok355);snf-11(ok156). Both N2 and snf-11(ok156) animals show very similar curvature pattern (Fig. 8B). In contrast, the smn-1(ok355) mutant animals display a distinct curvature shape. The normalized mean curvature in smn-1(ok355) animals decreases rapidly after the worm's mid-point and then sharply increases right near the tail. The smn-1(ok355);snf-11(ok156) mutant animals display a curvature pattern similar to smn-1(ok355) mutant animals but distinct from the N2 and snf-11(ok156) mutant animals, especially from the mid-point to tail region ( Fig. 8B; Extended Data Fig. 8-2; adjusted p Ͻ 0.00001 by multiple t tests and Holm-Sidak method correction). This result indicates that loss of the GABA uptake transporter (i.e., snf-11(ok156)) does not improve the curvature phenotype of the smn-1(ok355) mutant animals (Fig.  8B). These two means of enhancing GABAergic neurotransmission have non-identical effects on network behavior, a point that will be addressed further in Discussion.
Furthermore, we undertook a quantitative analysis of their kinematic and biomechanical properties (Fig. 9). Between L2 and late stage, all animals grew to a statistically significant increase in length (Fig. 9A). While N2 and snf-11(ok156) mutant animals display a progressive increase in swimming speed between L2 and late stage, the smn-1(ok355) mutant animals display a significant decrease in swimming speed. This decrement in swimming speed was prevented in the smn-1(ok355);oxls22  and smn-1(ok355);snf-11(ok156) mutant animals (Fig. 9B). There is no change in beating frequency between L2 and late stage in animals with genotype N2, oxls22  or snf-11(ok156), while beating frequency declines over this period in smn-1(ok355), smn-1(ok355);oxls22 , and smn-1(ok355);snf-11(ok156) mutant animals (Fig. 9C). In sum, these studies show that smn-1(ok355) animals have an age-dependent decline in swimming speed and beat frequency. Enhancing GABAergic neurotransmission in these animals selectively abrogates the swimming speed defect.
Next, we studied the propulsive force and mechanical power of these animals. With regard to propulsive force, only N2 and snf-11(ok156) mutant animals show an increase over the time (L2 vs late stage; Fig. 9D). With regard to mechanical power, N2, oxls22(UNC-49::GFP) and snf-11(ok156) animals display a progressively significant increase between L2 and late stage, while the smn-1(ok355) mutant animals do not display this agedependent change. In contrast, the smn-1(ok355); oxls22  and smn-1(ok355);snf-11(ok156) mutant animals do display a significant age-dependent increase in mechanical power (Fig. 9E). In sum, enhanced GABAergic neurotransmission has a selective effect on the dynamic biomechanical properties of the smn-1(ok355) mutant animals, it improves their mechanical power but not the propulsive force.

Discussion
Here, we show that reducing daf-2, insulin/IGF signaling (IIS) prolongs life and ameliorates aspects of motor dysfunction in C. elegans models of SMA. The effects of reduced daf-2 signal depend, at least in part, on the daf-16, a Forkhead box transcription factor. The daf-2/ daf-16 signaling pathway is conserved in metazoans, and is implicated in the control of human longevity and healthspan (Kojima et al., 2004;Suh et al., 2008). Some studies suggest that reduced insulin/IGF signal transduction contributes to the pathophysiology SMA (Hua et al., 2011), and provision of IGF1 is beneficial (Bosch-Marce et al., 2011;Shababi et al., 2011), while other work suggests that reducing levels of the IGF receptor is beneficial (Biondi et al., 2015). Precise control of the level and spatial-temporal regulation of IIS are likely to be critical determinants of efficacy. Here, our results demonstrated that the daf-2/daf-16 signaling pathway now joins this relatively small number of SMA disease phenotype modifiers, such as plastin 3 (Oprea et al., 2008), RhoA/ROCK (Coque et al., 2014), and four genes in C. elegans (ncbp-2, T02G5.3, grk-2, and flp-4;Dimitriadi et al., 2010).
In C. elegans, previous studies show the reduction of the IIS pathway in several neurodegenerative models by down regulation of daf-2 can suppress the shorten lifespan and ameliorate the motor deficit (Cohen et al., 2006;Florez-McClure et al., 2007;Li et al., 2013). Transgenic worms expressing human Aß 1-42 in the body wall muscles using the unc-54 promoter show aggregated-mediate toxicity and age-dependent motor deficit. These disease phenotypes were ameliorated by reducing the activity of the IIS pathway via daf-2 RNAi feeding (Cohen et al., 2006). Transgenic worms expressing Aß 1-42 in the muscles using the myo-3 promoter show an accumulation of autophagosomes, and this phenotype was alleviated by daf-2 mutations (Florez-McClure et al., 2007). In familial ALS models, crossing the daf-2(e1370) mutations into the transgenic worms expressing G85R (Boccitto et al., 2012) or G93A (Li et al., 2013) SOD1 mutations significantly ameliorated the motor deficits and the accumulation of insoluble aggregates. Likewise, transgenic worms overexpressing the full-length TDP-43 using a pan-neuronal snb-1 promoter resulted in an accumulation of insoluble aggregates and neurotoxicity, whereas daf-2 mutation reduced protein aggregates and alleviated disease phenotype . Consistent with the reports in other neurodegenerative models, here we provide the evidence that the reduction of daf-2 can improve the locomotion in C. elegans models of SMA. This observation is in line with work by Liu et al. (2013) showing that N2 worm locomotor function and neuromuscular transmission begin to decline in the second week of adulthood and this is blunted in the daf-2 mutant background. The extent to which these observations in middle aged adult worms are mechanistically relevant to our results on the smn-1 mutant animals will require further investigation.

The daf-2 pathway and locomotor defects in smn-1 mutant animals
The locomotor function of animals lacking smn-1 is comparable to wild type at early time points. However, normal developmental progression fails to occur in these animals and instead a severely uncoordinated phenotype emerges. The nervous system of smn-1(ok355) is, to a large degree, morphologically intact (Briese et al., 2009), although limited NMJ abnormalities do exist (Figs. 5, 6;O'Hern et al., 2017). A reasonable speculation for progressive motor deficits in smn-1 mutant animals is circuit dysfunction, that is, diminished SMN in neurons that drive motor activation are impaired in their operation. This supposition is consistent with observations made in Drosophila (Imlach et al., 2012) and mouse (Mentis et al., 2011) models of SMA. In fly, the reduced SMN state leads to many defects in larval neuromuscular function and all of these abnormalities were rescued by pan neuronal transgenic expression of SMN. Interestingly, restitution of SMN in fly brain excitatory cholinergic neurons (but not motor neurons or muscle) also rescues these abnormalities. Similar observations have been made in the SMA⌬7 mouse models (McGovern et al., 2015); overexpression of SMN in motor neurons alone has a minimal effect on survival and the animals are weak, while overexpression of SMN in motor neurons and other neurons had a much more robust survival promoting effect. Further, elimination Figure 9. Statistical analysis of biomechanical properties (BMP). Bar graphs represent BMP algorithm calculated (A) length (mm), (B) speed (mm/s), (C) frequency (Hz), (D) propulsive force (nN), and (E) mechanical power (pW) for the indicated genotypes from L2 or late stage. Data are shown as mean Ϯ SD; Ն10 animals per genotype were recorded. Nonparametric Kruskal-Wallis test followed by Dunn's multiple comparison test for significant; ‫ء‬ p Ͻ 0.05, ‫ءء‬ p Ͻ 0.01, ‫‪p‬ءءء‬ Ͻ 0.005, ‫‪p‬ءءءء‬ Ͻ 0.001, ns, no significance. A, Genotypes are well matched for length at the two time points. B, N2 and snf-11(ok156) animals show age-dependent increase while the smn-1(ok355) mutant animals show a decrease in speed. This decrease in speed is suppressed in the smn-1(ok355);oxls22(UNC-49::GFP) and smn-1(ok355);snf-11(ok156) mutant animals. C, Beat frequency remains constant over time in the N2, oxls22(UNC-49::GFP), and snf-11(ok156) mutant animals. It declines in the smn-1(ok355), smn-1(ok355);oxls22(UNC-49::GFP), and smn-1(ok355); snf-11(ok156) groups. D, Age-dependent increase in propulsive force is seen only in the N2 and snf-11(ok156) mutant animals. E, Age-dependent increase in mechanical power force is seen in N2, oxls22(UNC-49::GFP) and snf-11(ok156) but not in the smn-1(ok355) mutant animals. Age-dependent increase in mechanical power force is restored in the smn-1(ok355);oxls22(UNC-49::GFP) and smn-1(ok355);snf-11(ok156) mutant animals. of SMN from motor neurons has no effect on survival or strength, while elimination of SMN from both motor neurons and other neurons leads to a shortened life span and weakness. Together, these data suggest that reduced SMN in neurons that drive motor neuron activation is a major determinant of abnormal motor function in models of SMA.
What is the nature of the defect incurred by reduced SMN in premotor interneurons that leads to abnormal neuromuscular unit function? In the fly model of SMA, distinct cholinergic neuron populations contribute to distinct aspects of motor function, and this led Imlach et al., to conclude that ". . .the effect of SMN depletion on the motor network is an amalgam of specific defects in distinct neurons that sum to produce a generalize disruption of the motor system. . ." our observations complement this formulation. We find that (1) reduction in daf-2 signaling improves locomotion in the worm model of SMA, (2) this is associated with enhanced GABAergic neurotransmission, and (3) simply enhancing GABAergic transmission also improves locomotion. We propose that diminished GABAergic neurotransmission in the worm nervous system impairs the function of circuitry required for neuromuscular system operation in the smn-1(ok355) mutant animals. Consideration of the fly and worm data together supports the view that neurochemically defined populations of premotor interneurons are dysfunctional in the reduced SMN condition, and the sum of this interneuron dysfunction manifests as impaired motor behavior. It is noteworthy that the mouse (and likely human) spinal cord contains GABAergic and cholinergic subtypes of interneurons (Waldvogel et al., 1990;Caldeira et al., 2017). We suggest that manipulating the activity of subgroups of spinal cord interneurons could improve motor function in the reduced SMN condition.
GABAergic neurotransmission in C. elegans models of SMA C. elegans GABAergic neurotransmission is a fascinating topic. Classic work identifies 26 GABAergic neurons in the worm based on the criteria: (1) they contain GABA, (2) express the GABA biosynthetic enzyme glutamic acid decarboxylase (GAD,, and (3) express the GABA vesicular transporter (GAT, for loading GABA into synaptic vesicles (Gendrel et al., 2016). Recent work suggests this is an underestimate and at least two other neuron types could be involved in neuronal GABAergic transmission. One class of neurons, containing GABA and GABA uptake transporter (snf-11), have been dubbed "GABA clearance neurons" and a second class of neurons, containing GABA, GABA uptake transporter (snf-11) and GABA vesicular transporter (unc-47), have been dubbed "GABA recycling neurons." Based on their anatomy and position, both subtypes are poised to release GABA in addition to classical GABAergic neurons. Since 30 -40% of all synapses in the worm have GABA receptors, GABA may have both widespread and precise influence on circuit behavior (Gendrel et al., 2016).
We find that two different ways of enhancing GABAergic transmission in the smn-1(ok355) mutant animals led to improved locomotion function, strong evidence that improvement in mechanical power of the SMA mutant animals is a result of increased GABAergic tone. There are 7 known GABA receptor subtypes in the worm (Jorgensen, 2005;Gendrel et al., 2016) and the UNC-49 locus encodes three splice variants; the UNC-49::GFP worms we used here is an integrated line with GFP inserted in-frame with the B splice variant (Bamber et al., 1999). Part of GABA receptor diversity is thought to arise from alternative splicing of the UNC-49 locus in a cell typespecific manner. The existence of multiple copies of the UNC-49 locus in the UNC-49::GFP worms we used is likely to result in increased GABA receptor occupancy at synapses. However, since SMN functions in spliceosome assembly, the UNC-49::GFP;smn-1(ok355) mutant animals may have a quantitatively and qualitatively unique GABA receptor phenotype. As mentioned above, the snf-11 locus encodes the sole GABA uptake transporter in the worm and in the snf-11(ok156) mutant animals there is likely to be enhanced GABAergic neurotransmission owing to decreased GABA clearance. Considering the existence of GABA clearance and GABA recycling neurons, the snf-11(ok156) mutant animals may have a more widespread and perhaps less specific effect on interneuronal GABAergic communication. Although both mutant animals enhance the mechanical power of the smn-1(ok355) mutant animals, their different modes of action provide an explanation for why there is a restoration of body curvature when swimming in the smn-1(ok355);UNC-49::GFP, but not smn-1(ok355);snf-11(ok156), mutant animals.

Perturbation of the balance of excitatory to inhibitory (E/I) neurotransmission in C. elegans models of SMA
Our anatomic (Figs. 5, 6), functional (Fig. 7), and genetic (Figs. 8,9) studies are consistent with the view that an impairment of GABAergic neurotransmission is an important contributor to locomotor dysfunction in the smn-1(ok355) mutant animals. Studies of other neuronal populations come to complementary conclusions. For example, O'Hern et al. (2017) show that reduced levels of SMN leads to decreased microRNA-2 function, and this results in increased translation of GAR-2, the orthologue of the M2 muscarinic receptor (m2R). In mice, presynaptic M2R receptors control (i.e., inhibit) the time course of acetylcholine release from motor neurons (Slutsky et al., 2001;Parnas et al., 2005). The inference for the worm is that diminished SMN could enhance GAR-2 mediated inhibition of acetylcholine release and thereby impair the activation of cholinergic NMJs. However, since GAR-2 is expressed in a large number of neurons (Cao et al., 2017), increased GAR-2 expression is likely to be having a broad effect. Together with our observations, we suggest that an imbalance of E/I neurotransmission underlies, at least in part, the motor impairments in smn-1(ok355). This raises the possibility that pharmacological manipulation of these systems could be of benefit to individuals with SMA We note that the Hart lab reported smn-1(ok355) mutant animals to be resistant to aldicarb and levamisole (O'Hern et al., 2017), while we find smn-1(ok355) mutant animals switch from levamisole-resistant phenotype at L2 stage to both levamisole-and pyridostigmine-hypersensitive phenotype at the late stage. The origin of the disparity between these observations probably lies in differences in developmental stages, definition, and timing of paralyses, the Hart lab performed the functional tests at the L4 stage, while we performed the functional tests at the L2 and late (L2 ϩ 3 d, the fifth day after egg drop) stage.

Electrophysiological abnormalities in mouse models of SMA
Do observations in mouse models of SMA provide support for the view that anatomic/functional abnormalities in segmental spinal cord circuitry contribute to motor deficits? Study of the intrinsic properties of motor neurons and the monosynaptic reflex arc between proprioceptive afferents and motor neurons has revealed that SMA⌬7 motor neurons are hyperexcitable (Ling et al., 2010;Mentis et al., 2011). Despite this, stimulation of Ia afferents evokes smaller motor neuron potentials and this is likely to be due to either a reduced number of proprioceptive synapses on motor neurons, or a functional impairment of the remaining Ia-motor neuron synapses. This would be consistent with a specific reduction of VGlut1 expression on motor neuron soma in the SMA⌬7 mice (Ling et al., 2010;Mentis et al., 2011). There is only a small amount of data on inhibitory systems. McGovern et al. (2015) found that ablating SMN in GAD-expressing neurons had no phenotypic effects, and expressing SMN in GAD-expressing neurons in SMN⌬7 mice did not affect longevity. The extent to which these manipulations influence locomotion is unknown. Immunohistological studies reveal normal numbers of GABAergic (VGAT) and glycinergic (GlyT2) presynaptic terminals on motor neuron soma in the SMA⌬7 mice (Ling et al., 2010;Mentis et al., 2011). We do not know what effect reduced SMN has on the connectivity among segmental spinal cord interneurons, between segmental spinal cord interneurons and motor neurons, and the degree of functional alterations (i.e., patterns of activity) within these interneuronal populations. While technically challenging to study, we believe this is a critical gap in our understanding of why reduced SMN leads to weakness. Considering the remarkable molecular diversity of spinal cord interneurons (Bikoff et al., 2016), we predict selected subtypes of interneurons will be dysfunctional and/or aberrantly connected to partners in the reduced SMN state. These alterations could impair the activation of motor neurons leading to weakness.