Decreased microRNA levels lead to deleterious increases in neuronal M2 muscarinic receptors in Spinal Muscular Atrophy models

Spinal Muscular Atrophy (SMA) is caused by diminished Survival of Motor Neuron (SMN) protein, leading to neuromuscular junction (NMJ) dysfunction and spinal motor neuron (MN) loss. Here, we report that reduced SMN function impacts the action of a pertinent microRNA and its mRNA target in MNs. Loss of the C. elegans SMN ortholog, SMN-1, causes NMJ defects. We found that increased levels of the C. elegans Gemin3 ortholog, MEL-46, ameliorates these defects. Increased MEL-46 levels also restored perturbed microRNA (miR-2) function in smn-1(lf) animals. We determined that miR-2 regulates expression of the C. elegans M2 muscarinic receptor (m2R) ortholog, GAR-2. GAR-2 loss ameliorated smn-1(lf) and mel-46(lf) synaptic defects. In an SMA mouse model, m2R levels were increased and pharmacological inhibition of m2R rescued MN process defects. Collectively, these results suggest decreased SMN leads to defective microRNA function via MEL-46 misregulation, followed by increased m2R expression, and neuronal dysfunction in SMA. DOI: http://dx.doi.org/10.7554/eLife.20752.001


Introduction
Spinal Muscular Atrophy (SMA) is an autosomal recessive neurodegenerative disease and the leading genetic cause of infant death in the US (Cusin et al., 2003;Pearn, 1978). SMA is caused by homozygous deletion or mutation of the SMN1 (Survival Motor Neuron 1) gene, resulting in reduced Survival of Motor Neuron (SMN) protein levels (Lefebvre et al., 1995). SMN expression is ubiquitous, but particularly essential for motor neuron survival (Lefebvre et al., 1997). Disease severity, as well as spinal cord a-MN dysfunction and degeneration, correlates with the extent of SMN loss (Lefebvre et al., 1997). Understanding why SMN loss impairs function should offer insight into SMA and may reveal therapeutic targets. SMN is conserved across species (Miguel-Aliaga et al., 1999). Studies of various SMA models suggest a role for SMN in several cellular processes including snRNP assembly Yong et al., 2002), messenger RNA (mRNA) transport (Fallini et al., 2011), and local translation (Dimitriadi et al., 2010;Kye et al., 2014). SMN function, however, has not been linked definitively to MN degeneration or synaptic transmission defects caused by SMN loss. microRNAs (miRNAs) are non-coding RNAs that often repress protein translation, by a mechanism that requires miRNA binding to the 3'UTR of mRNA targets. Disruption of the miRNA pathway in spinal MNs leads to severe degeneration (Haramati et al., 2010). SMN loss alters levels and/or activity of specific miRNAs (Haramati et al., 2010;Kye et al., 2014;Valsecchi et al., 2015;Wang et al., 2014), but the cellular mechanisms leading to altered miRNA expression and/or function are unknown. The RNA helicase Gemin3 associates with both SMN and RNA-induced silencing complex components (Charroux et al., 1999;Hö ck et al., 2007;Hutvágner and Zamore, 2002;Meister et al., 2005;Mourelatos et al., 2002;Murashov et al., 2007). Gemin3 and SMN levels decrease concomitantly, suggestive of a functional link (Feng et al., 2005;Helmken et al., 2003).
We took advantage of the C. elegans SMA model to examine the connection between SMN, Gemin3, and miRNA function. SMN1, Gemin3, and multiple miRNA pathway components are conserved in C. elegans (Grishok et al., 2001;Miguel-Aliaga et al., 1999;Minasaki et al., 2009). Lossof-function (lf) mutations in smn-1, the C. elegans ortholog of SMN1, cause behavioral and morphological abnormalities, premature death, and sterility (Briese et al., 2009;Sleigh et al., 2011). smn-1 (lf) animals also have neuromuscular junction (NMJ) defects, suggesting a functional role for SMN-1 in MNs (Briese et al., 2009). MNs in smn-1(lf) animals do not die, likely because of their short lifespan. However, smn-1(lf) neuromuscular defects may correspond to the early stages of SMA pathogenesis, characterized by NMJ dysfunction prior to MN degeneration (Miguel-Aliaga et al., 1999;Yoshida et al., 2015). We find that the C. elegans Gemin3 ortholog, MEL-46, is perturbed by SMN-1 loss, impacting miR-2 suppression of the M2 muscarinic receptor ortholog, GAR-2 (Lee et al., 2000). Across species in SMA mouse models, we find decreased levels of miR-128, a potential miR-2 ortholog, and increased expression of the GAR-2 ortholog, m2R. Notably, m2R inhibition ameliorates axon outgrowth defects in MNs from a SMA mouse model, consistent with our results in C. elegans. eLife digest Spinal muscular atrophy is a genetic disease that causes muscles to gradually weaken. In people with the disease, the nerve cells that control the movement of muscles -called motor neurons -deteriorate over time, hindering the person's mobility and shortening their life expectancy. Spinal muscular atrophy is usually caused by genetic faults affecting a protein called SMN (which is short for "Survival of motor neuron") and recent research suggested that disrupting this protein alters the function of short pieces of genetic material called microRNAs. However, the precise role that microRNAs play in the disease and their connection to the SMN protein was not clear.
MicroRNAs interfere with the production of proteins by disrupting molecules called messenger RNAs, which are temporary strings of genetic code that carry the instructions for making protein. By disrupting messenger RNAs, microRNAs can delay or halt the production of specific proteins. This is an important part of the normal behavior of a cell, but disturbing the activity of microRNAs can lead to an unwanted rise or fall in crucial proteins.
O'Hern et al. made use of engineered nematode worms and mice that share genetic features with spinal muscular atrophy patients, including disruption of the gene responsible for producing the SMN protein. These animal models of the disease were used to examine the relationship between decreased SMN levels and microRNAs in motor neurons. The experiments showed that reduced SMN activity affects a specific microRNA, which in turn causes motor neurons to produce more of a protein called m2R. This protein is a receptor for a molecule, called acetylcholine, which motor neurons use to send signals to muscle cells.
Increased m2R may be detrimental to motor neurons. As such, O'Hern et al. decreased m2R protein activity to determine whether this could reverse the defects in motor neurons that arise in the animal models of the disease. Indeed, blocking this receptor rescued some of the defects seen in the animal models, supporting the link to spinal muscular atrophy.
Several treatments that block m2R are already available to treat other conditions. As such, the next step is to determine whether these existing treatments are able to protect mice models of spinal muscular atrophy against muscle deterioration or increase their lifespan. If successful, this could open new avenues for the development of treatments in people.
SMN-1 is required for normal NMJ function in C. elegans cholinergic MNs (Dimitriadi et al., 2016). Aldicarb is an acetylcholinesterase inhibitor that leads to acetylcholine accumulation in the NMJ and consequently, paralysis (Mahoney et al., 2006). The time course of aldicarb-induced paralysis was slowed by decreased SMN-1 activity (Dimitriadi et al., 2016). We tested if a decrease in MEL-46 function causes similar resistance to aldicarb and found that mel-46 loss of function resulted in aldicarb resistance across multiple alleles ( Figure 1B; Figure 1-figure supplement 1F and G), reminiscent of smn-1 loss. Reintroduction of mel-46 using the [mel-46(+)#1] rescue array restored aldicarb sensitivity in mel-46(tm1739) animals. Tissue-specific knock-down of mel-46 in cholinergic neurons resulted in aldicarb resistance, thus confirming that MEL-46 function is required in cholinergic neurons, as is SMN-1 ( Figure 1C) (Dimitriadi et al., 2016). We also showed that knock-down of mel-46 or smn-1 in inhibitory GABAergic neurons resulted in aldicarb hypersensitivity (Figure 1-figure supplement 1H). Our findings, taken together with previous work, suggest that MEL-46 and SMN-1 are required in both cholinergic and GABAergic neurons for normal NMJ function.
Perturbed MEL-46 (Gemin3) function likely contributes to synaptic defects in smn-1(lf) animals MEL-46 might act together with or downstream of SMN-1 in pathways necessary for NMJ function. To test these and other possibilities, we generated integrated multicopy transgenic lines expressing GFP-tagged MEL-46 expressed under control of the unc-17 cholinergic-specific promoter ( Figure 2A). MEL-46::GFP was found in both the cell bodies and processes of neurons.  Figure 1 continued on next page changes were seen in cytoplasmic MEL-46::GFP, leading us to evaluate localization of MEL-46::GFP in MN dorsal cord processes in smn-1(ok355) animals. Because ok355 deletion in smn-1 leads to a complete loss of function, smn-1(ok355) animals were maintained over an hT2 balancer and sterile smn-1(ok355) homozygous progeny carry some maternally-loaded SMN-1 protein (Briese et al., 2009). We found that MEL-46::GFP localizes to small granular structures in dorsal cord processes in control (smn-1(+)) and smn-1(ok355) animals. Our finding is consistent with previous work showing that Gemin3 localizes to granular structures in mammalian neurites; Gemin3 co-localizes with SMN in 50-60% of these granules, along with multiple mRNAs (Todd et al., 2010a(Todd et al., , 2010bZhang et al., 2006). In smn-1(ok355) animals, we found that the density of MEL-46::GFP-positive granular structures was doubled compared to smn-1(+) controls ( Figure  . These results suggest that decreased SMN-1 leads to MEL-46 mislocalization in cholinergic MN processes and diminished MEL-46 levels in granules. Our findings suggest that SMN-1 impairs MEL-46 function, which could contribute to smn-1(ok355) synaptic defects (Dimitriadi et al., 2016). To test this hypothesis, we increased mel-46 gene dosage in smn-1(ok355) animals using the [mel-46(+)#1] rescue array and showed that this ameliorated smn-1(ok355) aldicarb resistance defects ( Figure 2E). We also showed that increasing mel-46 specifically in cholinergic neurons, using the cholinergic-specific unc-17 (ACh) promoter in an integrated array, referred to as [ACh::mel-46::GFP], rescued smn-1(ok355) aldicarb resistance ( . It is possible that high levels of MEL-46 in cholinergic neurons cause aldicarb hypersensitivity, whereas broad overexpression of MEL-46 may impact NMJ function independent of cholinergic neurons. Taken together, our results suggest that loss of SMN-1 negatively impacts MEL-46 function, resulting in perturbed NMJ signaling. Our finding is consistent with observations in humans that reduced human SMN levels result in Gemin3 downregulation (Feng et al., 2005;Helmken et al., 2003),    The interdependency we report, between SMN-1 and MEL-46 levels, may be specific to particular tissues and/or neural circuits. For example, pharyngeal pumping rate defects were not ameliorated in smn-1(ok355) animals by increased mel-46 levels ([mel-46(+)#1] rescue array, Figure 2-figure supplement 1D), suggesting a privileged relationship between SMN-1 and MEL-46 in cholinergic NMJ signaling. Increasing mel-46 did rescue smn-1(ok355) synaptic protein localization defects. Using the [mel-46(+)#1] rescue array, we rescued smn-1(ok355) defective SNB-1 puncta width and intensity to normal levels, but did not ameliorate linear density defects ( Figure 2F-J). Notably, increased mel-46 in an smn-1(+) control background did not increase SNB-1 levels, suggesting that mel-46-induced up-regulation of SNB-1 is specifically beneficial in a smn-1(ok355) background. Increasing mel-46 with a second broadly-expressed mel-46 rescue array line, [mel-46(+)#2], also restored APT-4 puncta linear density to normal levels (Figure 2-figure supplement 1E), without rescuing other metrics. These results are consistent with the aldicarb/NMJ functional rescue studies presented here and suggest increasing mel-46 improves neuromuscular signaling in smn-1(ok355) animals by partially restoring levels and localization of synaptic proteins. Furthermore, these results suggest that mel-46 may act with or downstream of smn-1 in a pathway essential for NMJ function.

C.elegans miR-2 is required for NMJ function
The pathways in MNs downstream of SMN and Gemin3 that are linked to SMA are unknown. Here, we consider a role for these two proteins in miRNA regulation. As mammalian Gemin3 co-localizes and co-purifies with RISC pathway components (Hö ck et al., 2007;Hutvágner and Zamore, 2002;Meister et al., 2005;Mourelatos et al., 2002;Murashov et al., 2007), we considered a role for miRNA regulation in NMJ function. miRNA miR-2 is enriched in neurons, expressed at all developmental stages, and predicted to regulate expression of many proteins involved in neuronal development and function (Marco et al., 2012;Martinez et al., 2008). We hypothesized that miR-2 is necessary for proper NMJ function and that it might be perturbed by loss of either SMN-1 or MEL-46. To test this possibility, we first examined the aldicarb response of mir-2(lf) animals. Two different deletion alleles, gk259 and n4108, caused resistance to aldicarb paralysis compared to wild type animals ( Figure  . Taken together, we conclude that miR-2 is required in cholinergic neurons for proper NMJ signaling. As a first step towards evaluating how miR-2 loss impacts cholinergic MN presynaptic function, we examined the effects of miR-2 loss on localization of presynaptic proteins in DA MNs. Four fluorescently-tagged presynaptic proteins were examined: SNB-1 (synaptobrevin), SYD-2 (a-liprin), ITSN-1 (DAP160/Intersectin), and APT-4 (AP2 a-adaptin). Analysis of tagged SNB-1 in mir-2(gk259) animals revealed increased SNB-1 puncta linear density, but no change in puncta width or intensity compared to wild type animals ( Figure 3B-D). Additionally, mir-2(gk259) animals were indistinguishable from wild type control animals with respect to SYD-2 synaptic localization for all metrics analyzed (Figure 3-figure supplement 1C), suggesting that pre-synaptic active zones are unchanged in number and size; thus, synaptic changes are likely not the result of altered active zone number or size (Zhen and Jin, 1999). ITSN-1 puncta width and intensity, but not linear density, were decreased in mir-2(gk259) and mir-2(n4108), animals compared to wild type controls (Figure 3-figure supplement 1D-F). Finally, both mir-2(gk259) and mir-2(n4108) had decreased APT-4 puncta width, intensity, and linear density (Figure 3-figure supplement 1G-I). ITSN-1, similar to APT-4, is involved in vesicle recycling at the NMJ (Wang et al., 2008). Together with results from aldicarb resistance studies, our results suggest that loss of miR-2 results in synaptic dysfunction at the NMJ, consistent with decreased cholinergic synaptic release Sieburth et al., 2005). Additionally, we observed considerable overlap between synaptic protein localization defects resulting from miR-2 loss with those of smn-1(ok355) animals (Dimitriadi et al., 2016), a finding consistent with miR-2 and SMN-1 acting in partially redundant pathways at the NMJ.

C.elegans miR-2 targets gar-2 mRNA in cholinergic neurons
To address mechanistically how miR-2 loss impacts NMJ function, we searched for mRNA targets of miR-2. Canonically, miRNA loss results in overexpression of direct mRNA targets (Elbashir et al., 2001). Since miR-2 loss leads to aldicarb resistance, loss of the target(s) is expected to cause hypersensitivity to aldicarb. Following a literature search for genes whose loss of function results in hypersensitivity, we selected the following genes with putative miR-2 3'UTR binding sites for study: gar-2, dbl-1, sek-1, and vab-2 (Jan et al., 2011;Lewis et al., 2005;Paraskevopoulou et al., 2013;Reczko et al., 2012;Vashlishan et al., 2008). Loss of a bona fide target gene is predicted to suppress aldicarb resistance caused by miR-2 loss. Therefore, we crossed a deletion allele for each gene into the mir-2(gk259) background. Loss of any of these four genes suppressed mir-2(gk259) to some extent, but gar-2(ok520), which contains a large deletion removing gar-2 exons 6 and 7, resulted in the most complete suppression, thus suggesting that GAR-2 acts downstream of miR-2 ( Figure 4A;   . GAR-2 is a G protein-coupled acetylcholine receptor orthologous to the mammalian M2 muscarinic receptor (m2R) (Lee et al., 2000).
To determine if miR-2 regulates GAR-2 expression directly, we examined the consequences of perturbing the putative miR-2 binding site in the gar-2 3'UTR. Using CRISPR/Cas9-targeted genome editing, we scrambled the 18 base pair gar-2 3'UTR region corresponding to the endogenous miR-2 binding site ( Figure 4B). Compared to control animals (gar-2 UTRwt C ) carrying the disrupted PAM site mutation (C>T), we found that animals with the scrambled miR-2 binding site (gar-2 UTRscr C ) were resistant to aldicarb, similar to miR-2 loss ( Figure 4C). To evaluate whether disruption of the miR-2 binding site influenced gar-2 transcript levels, we compared gar-2 mRNA levels in gar-2 UTRscr C animals and gar-2 UTRwt C controls and found a 40% increase in gar-2 transcript in gar-2 UTRscr C ( Figure 4D). Disruption of the 3'UTR site likely inhibits binding of other miR-2 family members, possibly contributing to the effect we observe (Ibáñez-Ventoso et al., 2008).
To test the effects of miR-2 loss on GAR-2 function in cholinergic neurons, we undertook in vivo GFP reporter analysis of GAR-2 expression. We generated a construct encoding GFP with a gar-2 3'UTR, whose expression is driven under the control of a cholinergic-specific promoter (referred to as unc-17p-ACh::GFP::gar-2 3'UTRwt). A second control version of the construct contained the same scrambled UTR sequence as used in gar-2 UTRscr C animals (referred to as unc-17p-ACh::GFP::gar-2 3'UTRscr) ( Figure 4E). Transgenic lines were created by multicopy insertion for each construct. Increased GFP levels were observed in mir-2(gk259) animals expressing the intact 3'UTR construct as compared to control animals ( Figure  To quantify impact of miR-2 on expression of these GFP reporters in various genetic backgrounds, we compared relative changes in GFP levels of transgenic lines using a ratiometric strategy; we determined GFP expression for lines carrying unc-17p-ACh::GFP::gar-2 3'UTRwt and compared these values to lines carrying unc-17p-ACh::GFP::gar-2 3'UTRscr in the same genetic background (Figure 4-figure supplement 2D). By this method, we observed a~13% increase in relative GFP expression associated with loss of miR-2 ( Figure 4H), indicating that miR-2 directly suppresses translation of gar-2 mRNA by binding the gar-2 3'UTR at this 3'UTR site. We also assessed the effect of miR-2 loss on gar-2 transcript levels and found no significant difference in gar-2 mRNA levels between wild type animals and mir-2(gk259) loss of function animals ( Figure 4I). These results, combined with our reporter data, suggest that miR-2 binds and inhibits gar-2 mRNA translation, but does not reduce transcript levels. Previous studies have reported that miRNAs can influence protein synthesis of targets without destabilizing mRNA levels (Cloonan, 2015;Selbach et al., 2008).
Increased GAR-2 translation in animals lacking SMN-1 might be due to decreased mature miR-2 levels. To test this possibility, we used quantitative RT-PCR studies. After neuron-specific RNAi knock-down of either SMN-1 or MEL-46, we found decreases in mature miR-2 levels ( Figure 5B Figure 5C). This result is consistent with our finding that gar-2 transcript levels were unchanged in miR-2 complete loss conditions ( Figure 4I), despite alterations in GFP reporter expression ( Figure 4H). These results suggest neuronal miR-2 levels are decreased when MEL-46 or SMN-1 levels decrease.
Research in vertebrates has demonstrated m2R internalization normally occurs in response to chronic m2R stimulation, either by pharmacological agonist application or acetylcholinesterase inhibition (Clancy et al., 2007). Since endocytosis is defective in animals lacking SMN-1, perturbed endocytosis, in combination with miRNA misregulation, could contribute to GAR-2 accumulation at the membrane leading to decreased SNB-1 in smn-1(ok355) motor neurons (Dimitriadi et al., 2016). Loss of GAR-2 did not rescue smn-1(ok355) APT-4 puncta defects ( Figure 6-figure supplement  1C), but decreased APT-4 puncta width and intensity in smn-1(+) control animals. As loss of GAR-2 did not restore APT-4 synaptic defects, we conclude that there are additional pathways affected by smn-1 loss, beyond GAR-2 misregulation. Nevertheless, these results suggest that C. elegans GAR-2 levels are increased by smn-1 loss, which might contribute to NMJ defects in smn-1(lf) animals.
The GAR-2 mammalian ortholog, m2R, is increased in SMA mouse model motor neurons The closest human ortholog of GAR-2 is the M2 muscarinic receptor (m2R), encoded by the CHRM2 gene. GAR-2 and m2R are functionally conserved, as activation of these presynaptic receptors by acetylcholine in different species results in hyperpolarization and decreased NMJ acetylcholine release across species (Dittman and Kaplan, 2008;Dudel, 2007;Parnas et al., 2005;Slutsky et al., 2003). Previous research suggests decreased SMN function across species might impact miRNA activity across species, which could increase m2R levels consistent with our work in C.

Inhibition of m2R by methoctramine rescued axon outgrowth defects in SMA mouse model MNs
Decreased SMN levels results in axon outgrowth defects in MNs derived from a SMA mouse model (Rossoll et al., 2003) and increased m2R might contribute to this functional defect. To test this, we examined the impact of m2R pharmacological inhibition on axon length for DIV5 MNs from E13.5 wild type (FVB) and SMA mice (Smn -/-;SMN2 tg/0 ) ( Figure 7E). Wild type and SMA MNs were cultured in the presence of 50 nm or 500 nm methoctramine, an m2R antagonist. In wild type MNs, methoctramine decreased mean longest axon length. Conversely, methoctramine treatment in SMA MNs increased both mean longest axon length and total axon length ( Figure 7E; Figure 7-figure supplement 1A). We conclude that m2R inhibition rescues MN axon outgrowth defects in a SMA mouse model, consistent with a deleterious impact of increased m2R activity in SMA model MNs.
There is a paucity of research addressing the functional importance of Gemin3, in either the Gemin or the RISC complexes. Biochemical and co-localization studies support a role for Gemin3 in spliceosome assembly, mRNA transport, and miRNA function (Charroux et al., 1999;Dostie et al., 2003;Feng et al., 2005;Mourelatos et al., 2002;Zhang et al., 2006). It is possible that decreased and/or mislocalized Gemin3 impairs RISC function, leading to the diminished miR-2 activity and levels reported in Figure 5. However, since Gemin3 likely functions in numerous pathways, it may influence miR-2 through other, more indirect pathways. In Drosophila, Gemin3 is necessary for motor function, but the molecular mechanisms underlying this defect are unclear (Cauchi et al., 2008). Results presented here in C. elegans further support a requirement for Gemin3 in motor function and additionally, show that Gemin3 is capable of modifying miRNA levels and activity.
SMN depletion results in miRNA misregulation across species miR-2 belongs to the invertebrate K box family (motif: CUGUGAUA) of miRNAs (Ibáñez-Ventoso et al., 2008). It was suggested previously that miR-2 does not have well-conserved mammalian orthologs (Marco et al., 2012), but another study suggested that human miR-128 is a member of this miRNA family (Ibáñez-Ventoso et al., 2008). These two bioinformatics studies differ in their definition of the miRNA binding site, also known as the seed region. Our alignment of miR-2/ miR-128 miRNAs and gar-2/CHRM2 mRNAs suggests that the CUGUG seed region may be a conserved motif for miRNA binding ( Figure 7A). Both miR-2 and miR-128 are enriched in the nervous system and share conserved mRNA targets (Marco et al., 2012;Martinez et al., 2008). More studies will be needed to confirm that miR-2 and miR-128 are orthologs. Regardless, the results presented here, in conjunction with previous research, suggest that overall miRNA misregulation contributes to neuronal defects when SMN levels decrease (Haramati et al., 2010;Kye et al., 2014;Valsecchi et al., 2015;Wang et al., 2014;Wertz et al., 2016).
Altered function of miR-2/miR-128 or GAR-2/m2R is not sufficient to explain all of the dysfunction observed in models of SMN deficiency. In C. elegans, miR-2 loss does not cause overt defects and GAR-2 loss does not restore viability, fecundity, or normal development to animals lacking SMN-1 or MEL-46. This suggests miR-2 loss does not contribute to defects outside the NMJ caused by SMN-1 loss. And, at the NMJ, synaptic protein perturbations are more severe in animals with diminished SMN-1 or MEL-46, compared to those lacking either miR-2 or GAR-2, which is consistent with a broader range of defects caused by decreased SMN-1 levels. In mice, complete miR-128 knock-out results in decreased motor activity and premature death (Tan et al., 2013), but it is currently unknown how miR-128 loss might specifically impact cholinergic MN function.

M2 receptor expression is increased in SMA model MNs
We found that m2R levels were increased 50% overall in SMN-deficient mouse MNs compared to wild type by Western blot analysis ( Figure 7B and C). In C. elegans, relative GAR-2 translation was increased by only 5% globally when SMN-1 levels dropped, based on GFP reporter expression in cholinergic neurons ( Figure 5A). These two assays are not directly comparable, since the C. elegans experiment only investigated the contribution of miR-2 perturbation in regards to increased GAR-2 translation. Certainly, additional pathways are affected by SMN loss, beyond miRNAs, contributing to the greatly increased m2R levels in SMN-deficient MNs. These pathways may include mRNA transport metabolism, endocytosis, as well as spliceosome and ribonucleoprotein assembly (Dimitriadi et al., 2016;Fallini et al., 2011;Hosseinibarkooie et al., 2016;Pellizzoni et al., 2002;Yong et al., 2002). Extensive analysis would be required to understand how defects in these pathways contribute to increased m2R in SMN-deficient MNs.
Could increased M2 muscarinic receptor expression contribute to a-MN defects in SMA patients? m2R is expressed in a-MNs, with little to no expression in smaller gamma MNs (Welton et al., 1999). This expression profile correlates with the pattern of neurodegeneration in an SMA mouse model: selective loss of a-MNs, while gamma MNs remain unaffected (Powis and Gillingwater, 2016). Within a-MNs, m2R is distributed along the membrane and concentrates at postsynaptic connections with C-boutons (Deardorff et al., 2014). a-MNs in SMND7 mice have increased C-bouton sites (Tarabal et al., 2014), which could be an additional cause or consequence of increased m2R levels in a-MNs. Taken together, this evidence suggests that increased m2R is consistent with multiple features of SMA pathology. We also consider three additional previously defined m2R pathways as possible contributors to MN functional defects in SMN-deficient a-MNs: GIRK channels, Ca 2+ channels, and SK channels.
Classically, m2R receptor activation leads to GIRK-channel-dependent efflux of K + cations, resulting in neuronal hyperpolarization and decreased SV release (Sun et al., 2013). Therefore, increased m2R levels are consistent with the synaptic defects observed in SMA models across species (Dimitriadi et al., 2016;Kong et al., 2009). m2R activation of GIRK channels is conserved in C. elegans (Lee et al., 2000), suggesting GAR-2 loss may rescue NMJ defects in animals lacking SMN-1 by reducing GIRK channel activation. Interestingly, sustained GIRK activation has been previously linked to neurodegeneration (Coulson et al., 2008). m2R inhibits N-type Ca 2+ (Ca v 2.2) and P/Q-type Ca 2+ (Ca v 2.1) channels resulting in decreased SV release at the NMJ (Slutsky et al., 2003;Yan and Surmeier, 1996). And, Ca v 2.1-deficient mice exhibit NMJ degeneration and decreased active zone proteins (Fox et al., 2007;Nishimune et al., 2004). Additionally, in a mouse model of SMA, distal axons and growth cones had reduced Ca 2+ transients resulting from defective Ca v 2.2 excitability and accumulation (Jablonka et al., 2007). Increased m2R is consistent with decreased Ca 2+ channel activity observed in SMN-deficient animals. Moreover, Ca v 2.2 channels activate SK channels in a-MNs, suggesting increased m2R may lead to decreased SK channel currents (Goldberg and Wilson, 2005). The drug riluzole, which ameliorated motor defects in smn-1(ok355) animals, may act via SK channels (Dimitriadi et al., 2013). Increased m2R levels may result in excessive inhibition of SK channels, contributing to defective synaptic transmission in SMA models across species; this connection may offer additional mechanistic insight into the ameliorative effects of riluzole.
Previous reports suggest that decreases in Ca 2+ transients hinder axon outgrowth (Hutchins and Kalil, 2008). SMN loss also decreases these currents (Jablonka et al., 2007), consistent with defective axon outgrowth in SMN-deficient cultured neurons. Here, we show that inhibition of m2R by methoctramine ameliorates axon outgrowth defects in SMA mouse model MNs. As we find m2Rs are overexpressed in SMA MNs, methoctramine rescue of axon outgrowth may be the result of restored Ca 2+ channel function. Taken together, these data suggest that increased m2R expression contributes to axon outgrowth defects in SMA MNs and that m2R inhibition promotes axon outgrowth in SMA MNs.
We connect SMN functionally and mechanistically to the miRNA pathway. As an exemplar of this connection in two species, we demonstrate that decreased SMN levels lead to downregulation of specific miRNAs and consequent increased expression of M2 muscarinic receptors. Increased m2R activity is deleterious and consistent with a subset of the NMJ defects seen in SMA models, across species. We suggest future studies might address the possible benefits of m2R inhibition in SMA models, as a combinatorial approach with other therapies.

C. elegans strains, constructs and transgenes
Strains listed in Supplementary file 2A were maintained under standard conditions at 25˚C (Brenner, 1974); we provide complete genotypes with unique strain identifiers, consistent with the rigorous standards of the C. elegans community. Abbreviated names are sometimes used for arrays, integrated lines or alleles in Figures 1-5; additional information about abbreviations can be found in Supplementary file 2C. For experiments with smn-1(ok355) and smn-1(rt248), animals assayed were first generation progeny of hermaphrodites heterozygous for the hT2 balancer. To maintain a common genetic background, control smn-1(+) animals were also derived from +/hT2 parents. Similarly, for APT-4::GFP synaptic localization (Figure 2-figure supplement 1E) and aldicarb response studies ( Figure 6B), mel-46(tm1739) animals were first generation progeny of parents heterozygous for the nT1 balancer. Control mel-46(+) animals were derived from +/nT1 animals. For all other assays involving mel-46(tm1739), animals were first generation progeny of parents carrying the ytEx211 [mel-46(+)] rescue array; animals tested did not carry the array unless specified. For these experiments, N2 animals served as wild type controls. We attempted to generate smn-1(ok355);mel-46(tm1739) double mutant animals, but generation of heterozygous double mutant animals was not possible, using either balancer chromosomes or the ytEx211 [mel-46(+)] rescue array.

C. elegans behavioral assays
Aldicarb resistance assay: 1 mM aldicarb assays were completed in at least three independent trials blinded to genotype (n 30 animals/genotype) as described in previous work (Mahoney et al., 2006;Sato et al., 2009). Paralysis induced by aldicarb was scored as inability to move or pump in response to prodding with a platinum wire. Experiments involving smn-1(ok355), smn-1(rt248) or mel-46(tm1739) animals were completed at the early L4 stage. All other aldicarb experiments were done with young adult animals. Pharyngeal pumping: Assays were performed blinded to genotype as previously described (Dimitriadi et al., 2010). Pumping events were scored as grinder movement in any axis. Average pumping rates (± Standard Error of the Mean (SEM)) were pooled from at least two independent trials (n > 20 animals/genotype). Experiments involving smn-1(ok355), smn-1(rt248) or mel-46(tm1739) animals were completed at day three post-hatching (animals were kept at 25˚C for two days and then 20˚C for one day). Pumping experiments involving all other genotypes were done with young adult animals.

C. elegans light level microscopy
Animals were mounted on 2% agar pads and immobilized using 30 mg/mL BDM (Sigma) in M9 buffer. Dorsal cord protein localization: Images were obtained as Z-stacks of the dorsal cord above the posterior gonad reflex (100x objective, Zeiss (Jena, Germany) AxioImager ApoTome and Axiovision software v4.8). For MEL-46::GFP analysis, a set area was defined for each image along the dorsal cord (25 mm x 5 mm). Using ImageJ (RRID:SCR_003070), a uniform threshold was used to eliminate background. The number (density), mean fluorescence (intensity) and area (size) for MEL-46::GFP granular structures were calculated using the ImageJ 'particle analyzer' program. For synaptic protein localization, mean puncta width (meanfixedwidth), intensity (meanfixedvolume) and linear density (fixedwidthlineardensity) were quantified with an in-house developed program called 'Punctaanalyser' using MatLab software (v6.5; Mathworks, Inc., Natick, MA, USA; RRID:SCR_001622) (Kim et al., 2008). At least three independent trials (n > 17 animals/genotype) were performed. For data sets involving smn-1(ok355), smn-1(rt248), or mel-46(tm1739) animals, all genotypes were examined at the early L4 stage, while other data sets were collected with young adult stage animals. GFP Fluorescence Quantification: GFP images of L4 animals were acquired (10x objective, Zeiss V20 stereoscope and Axiovision software v4.8). Mean GFP fluorescence was quantified using ImageJ (RRID:SCR_003070). A threshold was set to eliminate background fluorescence. For each data set, thresholds were kept constant. Average fluorescence values (±SEM) were combined from at least three independent trials for n > 25 animals/genotype; however certain backgrounds containing rtEx855 [mel-46(+)] had a lower n (reported in legends) as these animals went sterile and/or did not throw many progeny carrying the mel-46 array. Ratios in Figure 4H and Figure 5A were calculated as average mean fluorescence for each genotype in the rtIs56 background and divided by their respective average mean fluorescence in the control rtIs57 background. Ratio SEM was calculated by summing the SEM for each population (see Figure 4-figure supplement 2D). All representative images shown were analyzed as part of data collection.
C. elegans total RNA isolation, cDNA synthesis and qPCR For each RNA sample, animals were synchronized by collecting eggs for 6 hr from gravid adults on large seeded NGM plates. After two days at 25˚C, young adult progeny were washed off, rinsed and flash frozen. Total RNA was extracted after a 15 min Trizol (Thermo Fisher, Waltham, Massachusetts) incubation. 1 ng total RNA was used for reverse transcription with either the miScript II RT kit (Qiagen #218160) for miRNA or the SuperScript III First-Strand Synthesis Supermix kit (Invitrogen #11752050) for mRNA. Methodology followed manufacturer's instructions. miRNA levels were determined in a 10 ml reaction using miScript SYBR Green PCR kit (#218073, Qiagen, Venlo, Netherlands) and 300 nM of mature miR-2 primer/probe. miR-60 was used to normalize miR-2 expression as it is not expressed in the nervous system where SMN-1 or MEL-46 were knocked-down. Forward primer sequences for miR-2 and miR-60 were, respectively: 5'-TATCACAGCCAGCTTTGATGTGC-3' and 5'-TATTTATGCACATTTTCTAGTTCA-3'. A universal reverse probe was provided by Qiagen. Primer sequences for act-1: 5'-acgccaacactgttctttcc-3' and 5'-gatgatcttgatcttcatggttga-3' (Ly et al., 2015). Primer sequences for 18S rRNA: 5'-TTGCTGCGGTTAAAAAGCTC'3' and 5'-CCAACCTCAAACCAG-CAAAT-3' (Essers et al., 2015). The stability of miR-60, 18S rRNA, and act-1 housekeeping RNAs were evaluated using the 'model-based approach to estimation of expression variation' (Andersen et al., 2004). mRNA levels were determined in a 10 ml reaction using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific # 4368706), and 300 nM of each primer. PGK-1 was used to normalize gar-2 expression, as the mammalian orthologue has been used previously as a housekeeping gene for experiments involving SMN (Abera et al., 2016;Simard et al., 2007). Primer sequences for gar-2: 5'-CCTGAACTCTCATTGCCCTTTATTGATGC-3' and 5'-CTAGCAGTCCC TTGCATTGAAAC-3'. Primer sequences for pgk-1: 5'-GGCCCTCGACAACCCAGCTCGTC-3' and 5'-CGGCGGAGGAATGGCCTATACC-3. All reactions were performed in triplicate. Melting curve analysis and electrophoresis in agarose gel of every PCR product was conducted after each qRT-PCR to control amplification specificity. Gene expression level was calculated as the fold change of relative DNA amount of a target gene in a target sample and a reference sample normalized to a reference gene using the comparative DDCT method as previously described (Kurrasch et al., 2004).

Statistical analysis
Log-rank test, two-tailed Mann-Whitney U-test, or t-test were used for C. elegans statistical analysis. The Mann Whitney U-test was chosen over t-test for experiments where homogeneity could not be assured (i.e. RNAi; extrachromosomal arrays; or potential maternal loading from a heterozygous parent). t-test was used to determine significance for spinal motor neuron Western blot quantification and qPCR quantification.