Patient-specific variants of NFU1/NFU-1 disrupt cholinergic signaling in a model of multiple mitochondrial dysfunctions syndrome 1

Multiple mitochondrial dysfunctions syndrome 1 (MMDS1) is a rare autosomal-recessive disorder caused by mutations in the protein NFU1. NFU1 is responsible for trafficking iron–sulfur clusters (ISCs) to recipient proteins within mitochondria that require ISCs for function (reviewed in Braymer and Lill, 2017; Maio et al., 2020; Rouault and Maio, 2017). Disruptions of ISC biogenesis or delivery result in significant negative effects on mitochondrial function, including impaired oxidative respiration and excessive oxidative stress (Maio and Rouault, 2020; Stehling and Lill, 2013; Vanlander and Coster, 2018; Wachnowsky et al., 2018). Consequently, pathogenic mutations of NFU1 cause severe disease. Tragically, MMDS1 is pediatric lethal in nearly every known case (Ahting et al., 2015; Ames et al., 2020; Bai and Kong, 2017; Birjiniuk et al., 2020; Cameron et al., 2011; Ferrer-Cortès et al., 2016; Invernizzi et al., 2014; Jin et al., 2017; Navarro-Sastre et al., 2011; Nizon et al., 2014; Seyda et al., 2001; Souza et al., 2018; Tonduti et al., 2015; Uzunhan et al., 2020).

Complicating our understanding of MMDS1 pathogenesis are the small number of cases (38 reported to date) and the diverse genetic variations of NFU1 that are potentially pathogenic. Twenty different NFU1 variants, including point mutations, frameshifts and small deletions, have been reported. Further, the NFU1 locus in approximately half of MMDS1 patients is compound heterozygous for different NFU1 variants. Most of these NFU1 alleles are thought to result in severe loss of NFU1 function. Consistent with this assumption, typical MMDS1 patient symptoms include neurological regression, reduced motor control (dystonia) and pulmonary hypertension (Ahting et al., 2015; Ames et al., 2020; Bai and Kong, 2017; Birjiniuk et al., 2020; Cameron et al., 2011; Ferrer-Cortès et al., 2016; Invernizzi et al., 2014; Jin et al., 2017; Navarro-Sastre et al., 2011; Neveu et al., 2022; Nizon et al., 2014; Seyda et al., 2001; Souza et al., 2018; Tonduti et al., 2015; Uzunhan et al., 2020). Neurological regression is likely to be caused by leukoencephalopathy, a lesion of white matter in the brain, and a common finding in reports of MMDS1 (Ames et al., 2020; Ferrer-Cortès et al., 2016; Invernizzi et al., 2014; Jin et al., 2017; Navarro-Sastre et al., 2011; Nizon et al., 2014; Souza et al., 2018; Tonduti et al., 2015). Pulmonary hypertension is likely to be caused by hyperproliferation of endothelial cells of the pulmonary vasculature (James et al., 2020; Niihori et al., 2020), but the underlying mechanism remains unclear. The mechanism underlying neuromuscular dysfunction in MMDS1 has not been directly explored.

Regulated control of movement depends on neuromuscular junctions, specialized synapses between motor neurons and skeletal muscle. Neurotransmitters secreted into this synapse drive either contraction or relaxation of the muscle. Skeletal muscle contraction is stimulated by the neurotransmitter acetylcholine (ACh). Cholinergic motor neurons secrete ACh into the neuromuscular junction, where it binds to either the fast-acting nicotinic ACh receptor (nAChR) or slower-acting muscarinic ACh receptor (mAChR). nAChRs are ligand-gated Na⁺ channels, whereas mAChRs are G-protein-coupled receptors that regulate second-messenger cascades (reviewed in Mukund and Subramaniam, 2020). The neurotransmitter γ-aminobutyric acid (GABA), secreted from GABAergic motor neurons, also acts at the neuromuscular junction, where it binds to the GABA receptor (GABAR) to antagonize muscle contraction. Like ACh receptors (AChRs), GABARs can be fast or slow acting. The fast-acting GABA_A receptor is also a ligand-gated ion channel and allows Cl⁻ to enter the muscle cell, thereby promoting muscle relaxation. The slow-acting GABA_B receptor relies on second-messenger cascades that also inhibit muscle contraction (Bettler et al., 2004; Ochoa-de la Paz et al., 2021; Shaye et al., 2021). Dysregulation of either of these signaling pathways can result in muscular phenotypes.

We previously established a Caenorhabditis elegans model of MMDS1 by recreating exact patient-specific NFU1 variants in the C. elegans ortholog NFU-1. We studied five variants affecting the ISC interaction domain of NFU1/NFU-1 and demonstrated that they caused dysregulation of cellular iron as well as oxidative stress (Kropp et al., 2021a). In agreement with biochemical studies from another research group (Wachnowsky et al., 2017; Wesley et al., 2017a,b), our data supported a role for altered NFU1/NFU-1 dimerization as the cause of allele-specific phenotypes of NFU-1 variants. Indeed, altered dimerization dynamics of NFU1/NFU-1 not only impairs ISC delivery, but can also expose ISCs to the intracellular environment, causing distinctive metallostress phenotypes. We thus hypothesized that these patient-specific variants would also result in unique neuromuscular phenotypes and could help elucidate the mechanism of neuromuscular dysfunction in MMDS1 patients. C. elegans has long been an excellent model organism for investigating neuromuscular function (Brenner, 1974), in no small part due to its exquisitely mapped neural connectome and gene expression atlas for the nervous system (Cook et al., 2019; Taylor et al., 2021) Both ACh and GABA signaling are well conserved at genetic and functional levels (Jones and Sattelle, 2004; Mongan et al., 1998; Kaupmann et al., 1997; Schofield et al., 1987), making the findings here relevant to human MMDS1.

In this study, we focused on two patient-specific NFU-1 variants: Gly147Arg and Gly166Cys. These two variants were chosen because (1) our previous work demonstrated that they have phenotypes that are divergent from a full loss of NFU-1 function, and (2) they are amongst the most common variants observed in both homozygous and compound heterozygous cases of MMDS1. These two alleles account for >50% of homozygous MMDS1 cases. We found that Gly147Arg and Gly166Cys result in allele-specific phenotypes, including dysregulation of ACh signaling, that differ from both wild type (WT), a full nfu-1 deletion (nfu-1Δ) and each other. Dysregulation does not appear to be caused by oxidative stress, but rather hypersecretion of ACh by Gly147Arg motor neurons in an apparent phenotypic gain-of-function and hypersensitivity to ACh of Gly166Cys muscle cells. Re-expression of WT nfu-1 in muscles was capable of rescuing motility phenotypes in both Gly147Arg as well as Gly166Cys, whereas rescue by re-expression of nfu-1 in neurons was more variable in both variants. Motility phenotypes of Gly147Arg were rescued by reducing secretion of ACh, indicating a potential avenue for therapeutic intervention in MMDS1 patients with this specific variant.

We previously demonstrated that patient-specific MMDS1 variants in the C. elegans gene nfu-1 display mitochondrial stress, including morphological and functional impairments, in multiple tissues (Kropp et al., 2021a). Because mitochondrial stress can result in movement defects due to dysfunction in the neuromuscular system, we assessed the motility of nfu-1 variants. We focused on the NFU-1 variants Gly147Arg and Gly166Cys (for clarity, these variants will be referenced by the corresponding amino acid change). The Gly147Arg and Gly166Cys variants were compared to the wild type (WT) and to a genetic deletion of nfu-1 (nfu-1Δ). Because Gly166Cys and nfu-1Δ arrest during late larval development (∼L4 larval stage), all analyses were performed at this stage regardless of genotype.

We used video recordings to quantify the motility of nfu-1 variants on solid medium. Videos of animals crawling on an agarose pad were captured and analyzed with the WormLab system (Restif et al., 2014). Gly166Cys and nfu-1Δ moved (either forward or backward) at significantly slower speeds than WT; however, Gly147Arg moved significantly faster than all other groups (Fig. 1A). In C. elegans, forward movement depends on a sinusoidal wave that begins at the nose and transits through the body to the tail. Aberrations in this waveform can be indicative of altered neuromuscular function. Both Gly166Cys and nfu-1Δ showed a decreased wavelength for sinusoidal movement in comparison to WT (Fig. 1B,D) that is indicative of a more compact wave. This phenotype was most severe for nfu-1Δ. All nfu-1 variants – Gly147Arg, Gly166Cys and nfu-1Δ – showed an increased wave amplitude (Fig. 1C,D), indicating deeper body bends.

We also analyzed turning and reversals, body movements normally observed during foraging behavior as animals search for food. Turning occurs when animals stop and change direction before continuing forward movement. Reversals occur when animals stop forward movement and instead move backward by initiating sinusoidal waves at the tail rather than at the nose. Gly147Arg turned more frequently than any other group, but no difference was observed between WT and Gly166Cys and nfu-1Δ (Fig. 1E). Further, each nfu-1 variant reversed more frequently than WT, with nfu-1Δ switching from forward to reverse movement more than any other group (Fig. 1F). This increased incidence of reversals resulted in Gly166Cys and nfu-1Δ spending more than 50% of the time reversing, significantly more than either WT or the Gly147Arg mutant (Fig. 1G). Despite an increased incidence of reversals, Gly147Arg did not spend more time reversing than WT, thus the increased reversal incidence could be a function of more rapid movement overall. Additionally, the decreased speeds of Gly166Cys and nfu-1Δ were not a result of the time reversing as speed was calculated from movement in both directions (Fig. 1A). Together, these data demonstrate that all nfu-1 variants show motility defects on solid medium, with the Gly147Arg variant displaying hyperactive behavior (e.g. more frequent turns, faster speed), whereas the Gly166Cys variant and nfu-1Δ were hypoactive (e.g. less frequent turns, slower speed). All variants showed alterations in body waveform (e.g. decreased wavelength and amplitude). Together, these changes suggested that neuromuscular function was disrupted in Gly147Arg, Gly166Cys and nfu-1Δ.

As an additional test of potential motility defects in nfu-1 variants, we analyzed swimming behavior. C. elegans placed in liquid will characteristically thrash back and forth in a sinusoidal movement called swimming. We captured videos of animals and used WormLab to analyze swimming. In agreement with the locomotor phenotypes observed on solid medium, Gly147Arg showed hyperactive movement (elevated wave initiation rate), whereas the Gly166Cys variant and nfu-1Δ were severely hypoactive compared to WT (reduced wave initiation rate) (Fig. 2A; Movies 1-4). These allele-specific effects were also observed for swimming speed and overall activity (Fig. 2B-D). Similar to the decreased wavelength on solid medium (Fig. 1B), Gly166Cys and nfu-1Δ showed significantly more stretch to individual waves (closeness of head and tail; Fig. 2E,H), indicating deeper curves of individual body waves. This effect was slightly greater in nfu-1Δ than in Gly166Cys. Yet, both Gly166Cys and nfu-1Δ spent significantly more time curling (Fig. 2F), a behavior that occurs normally but is exaggerated in these variants; nfu-1Δ demonstrated this phenotype most severely. As also observed for crawling on solid medium, C. elegans will occasionally reverse while swimming, with waveforms beginning at the tail rather than at the nose. This behavior is apparent as a change in the direction of body bend while swimming (see nfu-1Δ in Fig. 2H). Surprisingly, the Gly147Arg variant spent significantly less time reversing than any other group, whereas this difference was not observed for Gly147Arg on solid medium (Fig. 2G). The significance of this disparity is unclear, but may be attributable to distinct differences in the operation of movement circuits in crawling versus swimming behavior (Pierce-Shimomura et al., 2008; Vidal-Gadea et al., 2011). Collectively, the crawling and swimming phenotypes of these variants indicate that the Gly147Arg variant causes a gain-of-function in locomotion, whereas both the Gly166Cys variant and nfu-1Δ cause loss-of-function in locomotion. Importantly, the gain-of-function observed in Gly147Arg is exclusively a phenotypic effect as this allele is not genetically dominant (Fig. S1). Thus, the Gly147Arg variant is recessive but presents with a gain-of-function locomotion phenotype.

Given the motility defects observed in the nfu-1 variants, function of the neuromuscular junction was next assessed. In both C. elegans body wall muscle and human skeletal muscle, contraction is stimulated by ACh release from cholinergic motor neurons (Fig. 3A). Relaxation is induced by secretion of GABA from GABAergic motor neurons (Fig. 3A). The coordinated action of these neurotransmitters is essential for the contraction and relaxation cycles that drive the sinusoidal movement of C. elegans (Jorgensen and Nonet, 1995; Richmond and Jorgensen, 1999).

Neuromuscular signaling was first assessed by exposing animals to aldicarb, an inhibitor of acetylcholinesterase (AChE), which normally degrades ACh in the synapse (Fig. 3A). Treatment with aldicarb results in accumulation of ACh and paralysis due to hypercontraction (Mahoney et al., 2006). Surprisingly, all nfu-1 variants were hypersensitive to aldicarb (Fig. 3B; Table S1). Hypersensitivity to aldicarb could be caused by either enhanced sensitivity to ACh or an attenuated response to GABA. To distinguish between these models, we performed paralysis assays with drugs that directly target either cholinergic or GABAergic signaling via their respective receptor. Levamisole is an nAChR agonist that activates muscle nAChRs and, like aldicarb, causes hypercontraction and paralysis. In agreement with the hypothesis that nfu-1 variants are hypersensitive to ACh, each variant was also hypersensitive to levamisole, albeit to differing degrees (Fig. 3C; Table S1). Piperazine is an agonist of the GABA_A receptor, and thus exposure of C. elegans to this drug causes flaccid paralysis due to complete relaxation of the body wall muscles. Intriguingly, both Gly147Arg and Gly166Cys were resistant to piperazine, whereas nfu-1Δ was unaffected (Fig. 3D; Table S1). Together, these results suggested that both ACh and GABA signaling are dysfunctional in Gly147Arg and Gly166Cys. Further, these findings indicated that there must be a mechanistic difference between Gly166Cys and nfu-1Δ, at least with respect to GABA signaling.

We sought next to investigate whether neuromuscular signaling defects were caused by presynaptic (neuronal) or postsynaptic (muscular) effects. For example, hypersensitivity to ACh could be caused by elevated secretion of ACh, hypersensitivity of target nAChRs or a combination of both. Therefore, mutant C. elegans strains with either presynaptic or postsynaptic alterations in ACh and GABA signaling were analyzed with either aldicarb, levamisole or piperazine. The drug-induced paralysis curves with each of these mutants were then compared to that of nfu-1 variants. Strains tested were tom-1(ok285), rig-3(ok2156), unc-47(n2409) and unc-49(e407). TOM-1 is a syntaxin-binding protein that positively regulates ACh vesicular docking through interactions with syntaxins (Gracheva et al., 2006). RIG-3 is a cell-surface immunoglobulin-domain-containing protein that positively regulates the abundance of the ACR-16 subunit of muscle AChRs (Babu et al., 2011). UNC-47 is the vesicular GABA transporter and is essential for packaging of GABA into secretory vesicles (Vashlishan et al., 2008). UNC-49 is an essential GABA_A subunit of the GABAR (Weeks et al., 2018). These strains were generally hypersensitive to aldicarb and levamisole, as expected based on published data (Fig. S2, Table S2) (Babu et al., 2011; Bhardwaj et al., 2020; Gracheva et al., 2006; Vashlishan et al., 2008; Weeks et al., 2018). Only tom-1(ok285) and unc-49(e407) were resistant to piperazine (Fig. S2, Table S2). Because of the specificity of levamisole and piperazine to the receptor of their respective signaling pathway, we focused on these two drugs rather than aldicarb in further analyses. When comparing both levamisole and piperazine paralysis curves, the tom-1(ok285) and unc-49(e407) strains most closely resembled the nfu-1 variants, with tom-1(ok285) more similar to Gly147Arg and unc-49(e407) more similar to Gly166Cys and nfu-1Δ. Because TOM-1 is a presynaptic protein and UNC-49 is a postsynaptic protein, these data provided the first indication that Gly147Arg may cause a presynaptic defect and that Gly166Cys and nfu-1Δ may disrupt postsynaptic function.

To further investigate the presynaptic and postsynaptic roles of nfu-1, neuron-specific and muscle-specific nfu-1 knockouts were generated using a flippase (FLP)/flippase recognition target (FRT) system (Muñoz-Jiménez et al., 2017). Using CRISPR insertions, the endogenous nfu-1 locus was flanked by FRT sites (Fig. 4A). When combined with strains expressing FLP under control of a tissue-specific promotor, nfu-1 was excised early in development in those cells in which FLP was expressed. The strains generated were the body-wall-specific knockout AG618 [nfu-1(av246); bqSi294; bqSi495] and the neuron-specific knockout AG621 [nfu-1(av246); bqSi294; bqSi506]. For simplicity, these strains will be referred to as nfu-1^Δmuscle and as nfu-1^Δneurons, respectively. The FLP-expressing parent strains [BN503 (bqSi294; bqSi495) and BN507 (bqSi294; bqSi506), respectively] were included as controls; these strains will be referred to as FLP^muscle and FLP^neurons, respectively. All four strains were healthy with no obvious growth defects.

To confirm efficacy of the FLP, imaging of an internal fluorescent reporter and gene expression analysis were performed. The FLP-expressing strains all contained an internal transgenic reporter of recombination (Fig. S3A) (Muñoz-Jiménez et al., 2017). This reporter contains the mCherry sequence including stop codon flanked by FRT sites and the GFP sequence in that order. Cells that do not express FLP express mCherry but not GFP. Cells that express the FLP express GFP but not mCherry because the mCherry sequence and stop codon have been removed. Upon confocal imaging of the nfu-1^Δmuscle and nfu-1^Δneurons animals, GFP was observed in body wall muscle cells of nfu-1^Δmuscle and in neurons of nfu-1^Δneurons, as expected (Fig. 4B). To confirm nfu-1 excision, we used fluorescence-activated cell sorting (FACS) to isolate GFP⁺ cells (Fig. S3B,C) for mRNA extraction and measured nfu-1 expression by quantitative reverse transcriptase PCR (qRT-PCR). Because FLP is active in GFP⁺ cells, nfu-1 mRNA should be reduced in GFP⁺ cells from either nfu-1^Δmuscle or nfu-1^Δneurons animals. As expected, nfu-1 expression was undetectable in the body wall muscle (GFP⁺) cells of nfu-1^Δmuscle and undetectable in neurons (GFP⁺) of nfu-1^Δneurons. Additionally, nfu-1 expression was robust in all other (mCherry⁺) cells from each sample (Table S3). Therefore, we conclude that the FLP/FRT system efficiently excised nfu-1 in a tissue-specific manner.

The tissue-specific nfu-1 knockouts were subjected to levamisole and piperazine paralysis assays at the L4 larval stage as previously conducted for the nfu-1 variants. Intriguingly, nfu-1^Δmuscle was hypersensitive to levamisole but showed no change in sensitivity to piperazine, indicating that nfu-1 is required for normal ACh signaling in body wall muscle, but dispensable for normal GABA signaling (Fig. 4C,D; Table S4). Surprisingly, nfu-1^Δneuron showed no alterations in sensitivity to either levamisole or piperazine (Fig. 4E,F; Table S4), suggesting that nfu-1 is dispensable for motor neuron function in C. elegans. To confirm that motor neurons were included in the recombined cells of the nfu-1^Δneuron strain, we used qRT-PCR to measure expression of motor neuron-specific genes (unc-17, unc-3, cha-1, cho-1) in the GFP⁺ and mCherry⁺ samples from sorted nfu-1^Δneurons cells. Each of these genes was highly enriched in the GFP⁺ cells compared to the mCherry⁺ cells, indicating that the FLP was indeed effective in motor neurons (Table S4).

Swimming characteristics of the tissue-specific nfu-1 knockouts were also assessed. As with all other assays, analysis was initially performed at the L4 stage. Most swimming phenotypes observed in the nfu-1 variants were either absent or subtle at the L4 stage of the tissue-specific knockouts (Fig. S3D-I). However, we observed during routine culture that nfu-1^Δmuscle animals became progressively more uncoordinated with age. To quantify this effect, we analyzed the movement of day 1 adult (D1A) animals for swimming phenotypes. At D1A, many phenotypes were substantially more pronounced. The wave initiation rate was significantly reduced in nfu-1^Δmuscle and slightly, but significantly, elevated in nfu-1^Δneurons (Fig. 4G). Swimming speed and the activity index were likewise reduced in nfu-1^Δmuscle (Fig. 4H,I), yet these two phenotypes did not reach statistical significance in nfu-1^Δneurons. The stretch phenotype was slightly but significantly reduced in nfu-1^Δmuscle and nfu-1^Δneurons (Fig. 4J). Curling was significantly reduced in nfu-1^Δmuscle, with no change in nfu-1^Δneurons, whereas reverse swimming remained unchanged (Fig. 4K,L). These data show that the hypoactivity of nfu-1^Δmuscle correlates with results obtained with Gly166Cys and nfu-1Δ (Fig. 2), whereas the subtle hyperactivity phenotypes of nfu-1^Δneurons mirror our findings with Gly147Arg (Fig. 2). Collectively, the tissue-specific nfu-1 knockouts demonstrate that the necessity for nfu-1 is different in motor neurons and muscle cells, and that loss of nfu-1 activity in either cell type can produce unique movement phenotypes.

In a complementary approach to the tissue-specific knockouts, we transgenically re-expressed the WT nfu-1 sequence with tissue-specific promoters using the SKI LODGE system of single-copy transgene insertions (Fig. 5A) (Silva-García et al., 2019). As a proof of principle, we engineered both C. elegans nfu-1 (avIs286 [eft-3p::nfu-1::unc-54 3′UTR] V) and human NFU1 (avIs287 [eft-3p::NFU1::unc-54 3′UTR] V) for expression in all somatic cells (referred to as nfu-1^+soma and NFU1^+soma, respectively). Each of these strains was healthy and had largely normal motility, although slight but significant differences from WT in swimming speed (slightly increased in most transgenic lines) and stretch (slightly decreased in most transgenic lines) were observed (Table 1). We crossed these alleles into nfu-1Δ (nfu-1Δ; nfu-1^+soma and nfu-1Δ; NFU1^+soma) to determine whether each was capable of restoring normal motility. All motility phenotypes were significantly improved in nfu-1Δ; nfu-1^+soma, with every measure being rescued to near-WT levels (Table 1; Fig. S4). Although marginally less effective than nfu-1^+soma, motility phenotypes were partially rescued in nfu-1Δ; NFU1^+soma (Table 1; Fig. S4), confirming that NFU-1 is the functional ortholog of NFU1, but human NFU1 may not be fully functional when expressed in C. elegans. These results indicated that this transgenic system was viable for tissue-specific nfu-1 rescue experiments.

We next generated WT nfu-1 re-expression constructs to restore expression in neurons (avIs285 [rab-3p::nfu-1::rab-3 3′ UTR] V) and muscles (avIs284 [myo-3p::nfu-1::unc-54 3′ UTR] I) (referred to as nfu-1^+neurons and nfu-1^+muscle, respectively). These transgenic lines were healthy, but nfu-1^+muscle did have a slight reduction in wave initiation rate, and both lines had decreased stretch, similar to the somatic re-expression lines (Table 1). These two transgenes were crossed into the Gly147Arg and Gly166Cys variants, yielding the strains Gly147Arg; nfu-1^+neurons, Gly147Arg; nfu-1^+muscle, Gly166Cys; nfu-1^+neurons and Gly166Cys; nfu-1^+muscle. Swimming behavior of these strains was analyzed with WormLab.

Based upon the tissue-specific knockouts, we expected nfu-1^+neurons to more effectively rescue Gly147Arg than nfu-1^+muscle. However, re-expression of WT nfu-1 had a greater effect when expressed in muscles of Gly147Arg animals than in neurons (Fig. 5B-G, Table 1). Most notably, the wave initiation rate and activity index were restored to near-WT levels in Gly147Arg; nfu-1^+muscle (Table 1). There was no rescue of the reversal phenotype of Gly147Arg in Gly147Arg; nfu-1^+muscle. Rescue of motility phenotypes was weaker in Gly147Arg; nfu-1^+neurons. The wave initiation rate was not changed in Gly147Arg; nfu-1^+neurons compared to Gly147Arg alone, indicating that re-expression of WT nfu-1 in neurons is insufficient to rescue the increased motility of Gly147Arg animals. Surprisingly, the swimming speed was significantly elevated in both Gly147Arg; nfu-1^+muscle and Gly147Arg; nfu-1^+neurons compared to Gly147Arg animals alone. We do not have a satisfactory explanation for this finding at this time.

Data from the tissue-specific knockouts suggested that nfu-1^+muscle would likely most effectively rescue Gly166Cys motility phenotypes. Re-expression of WT nfu-1 in either the muscles or neurons of Gly166Cys animals significantly improved all aspects of motility (Fig. 5B-G, Table 1). However, the effects tended to be greater in Gly166Cys; nfu-1^+muscle than in Gly166Cys; nfu-1^+neurons animals, as was most obvious in the wave initiation rate, swimming speed and activity index (Fig. 5B-D, Table 1). The exception to this trend is curling, for which the rescue was more significant in Gly166Cys; nfu-1^+neurons than in Gly166Cys; nfu-1^+muscle (Fig. 5F, Table 1). Together, these data confirm that the necessity for nfu-1 differs between muscles and neurons and that muscles may be more sensitive than neurons to NFU-1 dysfunction.

Our testing of mutants that specifically impair either cholinergic or GABAergic signaling (Fig. S2, Table S2) indicated that either signaling pathway could cause the paralysis phenotypes observed in the nfu-1 variants. We thus chose to investigate these pathways more specifically. With dysfunctional GABA signaling, a gentle nose touch characteristically results in the ‘shrinker’ phenotype due to impaired relaxation of body wall muscle (Fig. 6A). In shrinkers, gentle nose touch results in contraction of anterior body muscles but does not stimulate backward movement. Thus, the animal shrinks, unlike WT, which simply reverses. The nfu-1 variants were tested for the shrinker phenotype, but, unlike the control unc-47(n2409), none showed any difference from WT (Fig. 6B). The nfu-1 variants did appear to pause, neither shrinking nor reversing, somewhat more frequently when touched, but this phenotype is likely to be due to altered sensation of physical touch (Kaplan and Horvitz, 1993; Perkins et al., 1986; Ware et al., 1975). Alterations in mechanosensation are the subject of ongoing work and beyond the scope of this study. However, the lack of a shrinker phenotype in the nfu-1 variants indicated that altered GABA signaling was not the cause of motility phenotypes. The piperazine resistance of Gly147Arg and Gly166Cys was likely to be caused by excessive ACh signaling, such that relaxation was impaired even in the presence of a GABAR agonist.

The paralysis phenotypes of the nfu-1 variants, and Gly147Arg in particular, are consistent with hypersensitivity to ACh. UNC-17 is the vesicular ACh transporter (Alfonso et al., 1993), and unc-17 knockdown should decrease ACh secretion due to reduced ACh in secretory vesicles. We first tested this approach with tom-1(ok285), which hypersecretes ACh (Gracheva et al., 2006). Because C. elegans neurons are relatively insensitive to RNA interference (RNAi), a procedure to concentrate the RNAi feeding bacteria was employed to enhance knockdown of unc-17. This procedure was effective at partially rescuing the tom-1(ok285) hypersensitivity to levamisole (Fig. S5, Table S5), validating this approach for knockdown of ACh secretion.

The nfu-1 variants were treated with the unc-17 RNAi protocol and then subjected to separate levamisole and piperazine paralysis assays. Knockdown of unc-17 had no effect on Gly166Cys or nfu-1Δ paralysis when the knockdowns were compared to the empty vector control (Fig. 6C,D), indicating that hypersecretion of ACh is not the cause of motor dysfunction in these variants. Surprisingly, RNAi treatment by itself caused a slight rightward shift in the piperazine paralysis curve of nfu-1Δ (Fig. 6D compared to Fig. 3D), which we hypothesize is due to the drugs used for RNAi [isopropyl-β-d-thiogalactopyranoside (IPTG) and carbenicillin] or an effect caused by feeding with HT115 bacteria rather than the standard OP50. However, unc-17 knockdown in Gly147Arg restored sensitivity to levamisole and resistance to piperazine to near-WT levels (Fig. 6C,D; Table S6), indicating that hypersecretion of ACh contributes to motor dysfunction in the Gly147Arg variant. The ability of unc-17 RNAi to rescue the swimming phenotypes of Gly147Arg was thus tested, and, as expected, nearly all metrics of hyperactivity were restored to WT levels (Fig. 6E-J). These results thus suggested that Gly147Arg enhanced release of ACh, which can be rescued by reducing ACh secretion.

We previously showed that treatment with N-acetyl-L-cysteine (NAC), an antioxidant, could extend the lifespan of Gly147Arg, indicating that oxidative stress contributed to the stress phenotypes of this mutant. Oxidative stress is known to affect neuromuscular signaling (Jia and Sieburth, 2021; Kim and Sieburth, 2018, 2019), so the possibility that oxidative stress contributed to the neuromuscular phenotypes of the nfu-1 variants was explored. Animals were raised on plates containing NAC or H₂0 (vehicle) until the L4 larval stage and then subjected to levamisole and piperazine paralysis assays. Treatment with NAC had no effect on nfu-1 variant sensitivity to levamisole or resistance to piperazine (Fig. S6A,B, Table S7). Surprisingly, WT animals treated with NAC were hypersensitive to levamisole without a statistically significant change in sensitivity to piperazine (Fig. S6A,B, Table S7). At the time of writing, we are unaware of any other reports of this effect in WT C. elegans, and the mechanism remains unknown. In all, our results demonstrated the tissue-specific effects of nfu-1 variants on ACh signaling.

Our prior results (Kropp et al., 2021a) and those presented here indicate that patient-specific variants of nfu-1, both of which are reduction- or loss-of-function alleles, result in neuromuscular dysfunction, with each nfu-1 variant resulting in distinct neuromuscular defects (Fig. 7). Both Gly147Arg and Gly166Cys are hypersensitive to the ACh agonists aldicarb and levamisole and resistant to the GABA agonist piperazine. However, they have opposite motility phenotypes: Gly147Arg is hyperactive, whereas Gly166Cys, like nfu-1Δ, is hypoactive. Therefore, Gly147Arg could be narrowly defined as having a gain-of-function locomotion phenotype. The mechanistic differences between these variants are very likely due to the specific differences in how these variants coordinated ISCs. We and others have previously shown that the Gly147Arg variant of NFU-1 fails to dimerize as effectively around ISCs, thereby impairing delivery to target proteins and causing excessive oxidative stress (Kropp et al., 2021a; Wesley et al., 2017 a; Wesley et al., 2017b). Gly166Cys, owing to the additional cysteine residue, fails to release ISCs, thereby also impairing delivery to target proteins (Kropp et al., 2021a; Wachnowsky et al., 2017). Thus, although ISC delivery is disrupted in both nfu-1 alleles, the biological defects differ depending on the specific variant. One explanation could depend on different downstream effects on metabolic function. We previously demonstrated that Gly147Arg shows reduced function of the lipoic acid synthase (human, LIAS; C. elegans, LIAS-1), an essential metabolic enzyme and target of NFU1/NFU-1, but to a level sufficient to maintain normal respiratory capacity (Kropp et al., 2021a). In contrast, LIAS-1 function is largely eliminated in Gly166Cys and nfu-1Δ, resulting in severely reduced respiratory capacity. This severe loss of respiratory capacity likely impacts muscle cells in a way that the partial reduction of Gly147Arg does not.

Our results showing that RNAi knockdown of the ACh vesicular transporter, UNC-17, suppresses the hypersensitivity of Gly147Arg to levamisole suggest that Gly147Arg enhances the activity of cholinergic signaling, potentially by elevated ACh secretion. As such, Gly147Arg, which is hypersensitive to aldicarb, corresponds to a ‘Hic’ (hypersensitivity to inhibitors of cholinesterase) mutant in line with mutants causing hypersecretion of ACh (Lackner et al., 1999; Miller et al., 1999; Miller et al., 2000; Nurrish et al., 1999; Reynolds et al., 2005; Schade et al., 2005). This interpretation is consistent with ubiquitous nfu-1 expression in the nervous system, including the ventral nerve cord (Taylor et al., 2021). Other Hic genes are also expressed in the ventral nerve cord, where they act through regulation of diacyl glycerol to stimulate ACh secretion (Lackner et al., 1999; Miller et al., 1999; Miller et al., 2000; Reynolds et al., 2005). That most motility phenotypes were not rescued in Gly147Arg; nfu-1^+neurons complicates our understanding of the role of Gly147Arg in motor neurons despite the strong similarity to Hic phenotypes. It is possible that expression of WT nfu-1 was too low to overcome the effects of Gly147Arg. As noted above, the molecular mechanism causing potential ACh hypersecretion from neurons in Gly147Arg remains to be discovered, but it likely differs from that of either Gly166Cys or nfu-1Δ, which appears to alter the cholinergic response of body wall muscle.

Notably, a recent RNAi screen determined that knockdown of genes involved in ATP synthesis pathways result in hypersensitivity to levamisole (Chaya et al., 2021). These results indicate that disruption of mitochondrial metabolism in body wall muscle is sufficient to induce levamisole hypersensitivity. Our finding that re-expression of WT nfu-1 in the muscles of both Gly147Arg and Gly166Cys animals rescued many motility phenotypes supports this claim. Furthermore, it is likely that the levamisole hypersensitivity phenotypes of Gly166Cys and nfu-1Δ are caused by disrupted mitochondrial metabolism in body wall muscle cells given the nearly complete rescue of motility in the Gly166Cys; nfu-1^+muscle animals.

Another important consideration for these results is the possibility of neurodegeneration contributing to the observed neuromuscular defects. It is well documented in both humans and C. elegans that mitochondrial dysfunction including MMDS1 can result in neurodegeneration. Although the integrity of motor neurons was not investigated in this study, such analyses are well established in C. elegans (Caldwell et al., 2020; Kropp et al., 2021b) and remain a subject of interest for future work. An intriguing possibility as a source of neurodegeneration in the nfu-1 variants is ferroptosis, a unique cause of cell death caused by iron toxicity. Our previous work demonstrated that cellular iron is elevated in nfu-1 variants (Kropp et al., 2021a). Ferroptosis was ruled out as a cause of cell death in the germline, but its potential impact on neurons remains to be evaluated.

All nfu-1 variants displayed atypical movement. Gly147Arg animals demonstrated hyperactivity on solid medium and in liquid, as indicated by increased rate and speed of crawling or swimming, respectively. As demonstrated by phenotypic rescue by RNAi knockdown of the vesicular ACh transporter, unc-17, movement defects of Gly147Arg mutants are likely to be at least partially attributable to hypersecretion of ACh. Exceptions are stretch and time spent curling or reversing, which are not suppressed by unc-17 RNAi knockdown. Interestingly, neither stretch nor curling was different from WT in basal conditions, yet unc-17 RNAi significantly elevated both stretch and curling in Gly147Arg compared to empty vector RNAi. In fact, unc-17 RNAi seems to cause increased stretch regardless of genotype. However, curling while swimming is not a well-characterized phenotype, and thus the exact meaning of the increase with unc-17 RNAi is unclear. As reversals are generally considered an aspect of foraging behavior (see below), it is possible that the reduced reversals while swimming observed with Gly147Arg are not a primary phenotype of neuromuscular dysfunction and thus cannot be rescued by reducing ACh signaling. Therefore, these data demonstrate that many, but not all, of the Gly147Arg motility phenotypes are caused by ACh hypersecretion.

On solid medium and in liquid, Gly166Cys and nfu-1Δ were hypoactive, with multiple atypical patterns of movement. Both variants demonstrated dramatic decreases in wavelength with commensurate increases in amplitude on solid medium, indicating that each body wave was deeper and shorter than those of WT. In liquid, this phenotype appears to have manifested as increased stretch, meaning that each wave was deeper, bringing the head and tail closer together. The cause of this stretch phenotype remains unclear, but could be due to an alteration in the coordinated contraction and relaxation necessary for normal sinusoidal waves. Failure of these variants to relax fully due to a state of constant contraction could explain the hypersensitivity to levamisole, deeper bends and increase in curling observed here.

An unexpected finding was the dramatic increase in reversals of Gly166Cys and nfu-1Δ animals when crawling on a solid medium. Reversing is a normal aspect of foraging behavior (also stated above), especially during local searching for food (Calhoun et al., 2014; Gray et al., 2005; Pierce-Shimomura et al., 1999). Reversal behavior is generally considered to be a function of behavioral response to physical or chemical stimuli, which must be integrated to determine whether animals proceed forward, turn, or reverse and change direction (Gray et al., 2005; Pierce-Shimomura et al., 1999; Zariwala et al., 2003; Zhao et al., 2003). When removed from food, C. elegans will explore for food in part by chemotaxing toward attractive stimuli including olfactory signals (de Bono and Bargmann, 1998; Fujiwara et al., 2002; Hills, 2004; Sawin et al., 2000). When attractive stimulation is increasing, the rate of reversals is reduced. Thus, the increased incidence of reversals as well as duration of reversals in the nfu-1 variants, and Gly166Cys and nfu-1Δ in particular, suggest a chemosensory defect that causes this behavior. This phenotype is the subject of ongoing investigation.

Given the ability of antioxidants to extend the lifespan of Gly147Arg (Kropp et al., 2021a), we performed experiments to determine whether oxidative stress contributed to Gly147Arg motility phenotypes. Work in other systems has demonstrated that oxidative stress can negatively affect the function of AChRs, especially in neurodegenerative diseases such as Alzheimer's disease (Cacanyiova et al., 2013; Campanucci et al., 2008; Guan, 2008; Krishnaswamy and Cooper, 2012). Conversely, the stress (including oxidative) regulator SKN-1 can negatively regulate ACh release (Staab et al., 2013; Staab et al., 2014). This mechanism likely does not contribute to the cholinergic hypersensitivity of nfu-1 mutants given that nfu-1 RNAi activates SKN-1 (Kropp et al., 2021a), which would be expected to reduce ACh secretion. The most surprising finding was that the antioxidant, NAC, significantly increased the levamisole sensitivity of WT animals (Fig. 6A). At the time of writing, we are unaware of any previous reports of this finding. However, there is evidence that NAC can increase the contractility of skeletal muscle in other systems (Okay et al., 2010), potentially through inhibition of AChE (Abdel-Wahab and Moussa, 2019; Costa et al., 2015). Such inhibition would potentiate cholinergic signals, thereby increasing sensitivity to levamisole. Perhaps WT animals are sensitive to this level of regulation whereas the hypersensitivity of the nfu-1 variants masks any effect of NAC.

As noted above, dystonia is a shared phenotype of nearly all MMDS1 patients, so this work is applicable to human disease biology. The most significant finding is that the motility phenotypes of Gly147Arg can be rescued by reducing ACh signaling. Importantly, one MMDS1 patient is homozygous for Gly189Arg, the NFU1 orthologous residue of Gly147Arg modeled here. This individual presented with hypertonia rather than hypotonia, making the results here directly translatable from C. elegans to humans (Uzunhan et al., 2020). Furthermore, the hypertonia of this patient strengthens the conclusion that Gly147Arg/Gly189Arg, despite having reduced ISC delivery function, could result in elevated ACh secretion from motor neurons. This individual is also one of the few MMDS1 patients alive at the time of the case report, further indicating that Gly147Arg/Gly189Arg is a comparatively less severe variant than other documented NFU-1/NFU1 variants. These results suggest that a cholinergic antagonist could be relevant as a therapeutic intervention for MMDS1 patients with the Gly189Arg variant.

In contrast to the Gly147Arg variant, the phenotypic effects of the Gly166Cys variant are more consistent with a full loss-of-function of NFU-1 activity. This variant is not only the most common allele documented in MMDS1, but it is also more similar to other common NFU1 variants than Gly147Arg (Kropp et al., 2021a). These data, combined with data from nfu-1Δ and Gly166Cys, strongly suggest that other complete loss-of-function variants have reduced capacity to respond to cholinergic signaling, thereby causing hypotonia. As noted above, we chose to analyze these variants in isolation as homozygotes to best investigate the unique characteristics of each variant. However, approximately half of MMDS1 patients are compound heterozygous for multiple variants. As demonstrated in one recent study of endolysosomal trafficking dysfunction (van der Welle et al., 2021), analysis of compound heterozygous variants is highly effective in C. elegans. Such an approach would be possible with nfu-1 variants to better elucidate the genotype/phenotype relationship of individual MMDS1 patients. Although more research is needed to understand the molecular mechanisms of all NFU1/NFU-1 variants causing neuromuscular dysfunction, this work has identified cholinergic signaling as a prime target and developed tools for its investigation.

Animals were maintained following standard procedures (Brenner, 1974) on MYOB plates seeded with OP50 bacteria. All nfu-1 variants were maintained as heterozygotes balanced with tmC5. Any modifications to this process as required for experiments are described below. All strains used are listed in Table S8.

CRISPR gene editing was carried out as in Kropp et al. (2021a). In brief, CRISPR injection mixes were prepared and carried out by injection into the syncytial gonad with dpy-10 as a co-injection marker as described in Paix et al. (2017). Guide RNAs were purchased from Horizon Discovery, and repair sequences were either purchased from IDT or generated by PCR using Phusion polymerase (NEB). FRT sites around the endogenous nfu-1 locus were added sequentially with the 5′ sites added first. Owing to linkage of the nfu-1 locus with the bqSi495 and bqSi506 insertion site on chromosome IV, the FRT sites were inserted into each strain independently, resulting in the identical nfu-1(av246) allele in each genetic background. For transgenic rescue constructs, the SKI LODGE system of single-copy insertions was utilized (Silva-García et al., 2019). All C. elegans nfu-1 constructs utilized the entire nfu-1 genomic locus (exons and introns), whereas only the cDNA was inserted for the human NFU1 rescue construct. Injection mixes were prepared as described in Silva-García et al. (2019), with repair constructs at a concentration of ∼500 ng/μl in the injection mix. The nfu-1 or NFU1 sequences were first inserted into the appropriate tissue-specific driver strain (WBM1126, WBM1140, WBM1456). All CRISPR constructs were sequenced for accuracy, and at least two lines were independently generated for each allele. Alleles containing the tissue-specific nfu-1 or NFU1 rescue constructs were subsequently crossed into the appropriate nfu-1 variant, and transgenes were homozygosed prior to analysis. Guide RNA, repair sequences and primers for generation of repair sequences are listed in Table S9.

Videos were captured on a Nikon SMZ1245 microscope equipped with a Hamamatsu Orca Flash 4.0 LT camera controlled by Nikon Elements BR software. Videos were 1  min in length and captured at 33 frames/s for swimming or 7.5 frames/s for crawling. Videos were written and saved as MP4 files.

For swimming analysis, sample preparation included picking single animals from a culture plate into a watch glass containing M9 buffer to remove excess bacteria. Animals were then picked from the watch glass to a droplet of M9 buffer on a microscope slide and imaged immediately as described above. One or two animals were imaged per individual video.

For crawling analysis, around five animals were picked directly from a culture plate to an agarose pad. Excess bacteria were removed with a platinum pick.

Confocal microscopy was carried out using a Nikon Eclipse Ti2 microscope equipped with a Yokagawa CSUX-1 spinning disk and Photometrics Prime95B camera using Nikon NiS Elements acquisition software. Images were captured in widefield at 10× magnification.

Induction of fluorescent reporters was achieved by a 30-min heat shock of L4 samples at 34°C followed by 4 h of recovery at 20°C. Live imaging was carried out by mounting samples on 2% agarose pads and anesthetized with 5 mM levamisole. Samples were imaged through a #1 cover glass sealed with nail polish.

Movement phenotypes were assessed with WormLab software purchased from MBF Biosystems (Restif et al., 2014). MP4 files were imported and adjusted for accurate identification of animals. Scale was either 4.81 or 2.18 µm/pixel, and threshold was either 215 or 195 depending on zoom used. Background smoothing was set to 10 for all samples. Width fitting and backtracking were enabled. Following tracking, each track was manually screened for accuracy and repaired (switching head or tail) if the head of the animal was misidentified.

Animals for shrinker analysis were gently touched on the nose with an eyelash. Responses were scored as either a reversal (whole body reversal), shrinker (contraction of body without movement) or pause (stop in motion without shrinking or reversal).

C. elegans were treated with RNAi from the updated Ahringer library (Fraser et al., 2000) expressed in HT115(DE3) bacteria. To increase RNAi expression in neurons, a protocol modified from Timmons et al. (2001) was used to concentrate bacteria. RNAi cultures were concentrated as follows: 3 ml LB [100 mg/ml ampicillin (amp)] medium inoculated with RNAi or empty vector bacterial clone. Cultures were grown for 16 h at 37°C in a drum roller. Then, 100 µl of this culture was inoculated into 5 ml LB (100 mg/ml amp) and incubated for 4 h at 37°C in a drum roller. Following this incubation, IPTG was added to a final concentration of 1 mM, and cultures were incubated for another 4 h at 37°C in a drum roller. Following this incubation, cultures were centrifuged for 10 min at 3750 g in a benchtop centrifuge. The supernatants were aspirated, and bacterial pellets were resuspended in 2 ml LB (100 mg/ml amp) medium. This culture was seeded onto MYOB plates containing 4 mM IPTG and 100 mg/ml amp. Bacteria were allowed to dry for 48 h prior to use. All plates and cultures were prepared fresh for each experiment.

L4 hermaphrodites were plated onto RNAi plates prepared as above. After 24 h, these adult animals (P₀) were plated onto a fresh RNAi plate for a 2-h egg lay. P₀ animals were removed after 2 h, and embryos were allowed to hatch and develop. Animals were used for experiments when they reached the L4 stage, which varied depending on genotype. Data were accumulated from a minimum of three independent experiments.

MYOB plates containing 5 mM NAC were prepared fresh for each experiment. NAC was added immediately prior to pouring. Volume-matched vehicle plates (0.8% H₂O) were simultaneously prepared. Plates were seeded with OP50 bacteria and allowed to dry for 48 h prior to use. L1-arrested larvae (see below) were plated onto NAC or vehicle plates and allowed to grow to the L4 stage, at which point they were used in paralysis assays. Data were accumulated from a minimum of three independent experiments.

Arrested L1 larvae were obtained through sodium hypochlorite treatment of gravid adult hermaphrodite animals and subsequent incubation in S-basal buffer overnight.

Paralysis assays were all performed over 4 h with timepoints every 15 or 20 min depending on assay. Animals were plated at time 0, and assessed for paralysis at the indicated timepoints by touching three times (midbody, head, tail) with an eyelash pick. Animals were considered paralyzed when no response occurred from all three touches.

Drug plates were prepared the day before each assay. Drugs were added to liquid MYOB medium immediately prior to pouring. The final concentration of each drug was as follows: aldicarb, 1 mM (stock in 70% EtOH); levamisole, 40µM (stock in in H₂O); piperazine, 150 mM (stock in 100% EtOH). Data were accumulated from a minimum of three independent experiments.

To generate synchronized cultures, embryos were obtained by hypochlorite treatment of adult hermaphrodites and allowed to hatch in M9 buffer overnight (16-23 h). L1-arrested larvae were then grown on 150 mm 8P (peptone-enriched) plates seeded with NA22 bacteria for 39-43 h at 23°C to promote rapid expansion of the population (Spencer et al., 2014; Taylor et al., 2021; Zhang et al., 2011). For the heat-shock treatment, the 8P plates were placed in a single layer in a 34°C incubator for 30 min and then transferred to a 20°C incubator to recover for 3 h. After treatment, populations were L4 larvae and young adults.

Cell dissociation and FACS were conducted as previously described (Spencer et al., 2014; Taylor et al., 2021; Zhang et al., 2011). Briefly, worms were washed from the plates and separated from the bacteria through slow speed spins in cold M9, yielding two ∼300 µl wet pellets of animals. Pellets were transferred to 1.7 ml tubes and treated with 2× volumes of SDS-DTT (20 mM HEPES, 0.25% SDS, 200 mM DTT, 3% sucrose, pH 8.0). Egg buffer was added to stop the reaction. The worms were washed in 1.0-1.5 ml egg buffer, resuspended in 2× volumes of 15 mg/ml pronase and dissociated by pipetting. The pronase digestion was stopped with 1.0-1.5 ml L-15-10 (Gibco L-15 medium, 10% fetal bovine serum, 0.5% penicillin/streptomycin). After centrifugation, cells were resuspended in 1 ml cold egg buffer and separated from large debris with a low-speed spin (100 rcf for 2.5 min at 4°C). The supernatant was passed through a 35-μm filter into a collection tube. The pellet from the low-speed spin was resuspended and re-pelleted at low speed to recover additional cells in the supernatant, which were passed through the 35-µm filter and added to the collection tube.

FACS was performed on the cell suspension in the collection tube on a BD FACSAria™ III equipped with a 70-μm-diameter nozzle. Then, 4′,6-diamidino-2-phenylindole (DAPI; 1 mg/ml) was added to the sample (final concentration of 1 µg/ml) to label dead and dying cells. Cells were separated for either GFP or mCherry expression. Sorted cells were collected directly into Trizol LS and stored at −80°C for RNA extractions.

RNA extraction, cDNA synthesis and qRT-PCR were carried out as in Kropp et al. (2021a). In brief, RNA was extracted in Trizol Reagent (Thermo Fisher Scientific) and purified with a Zymo QuickRNA MiniPrep kit (Zymo). Synthesis of cDNA was carried out with iScript cDNA Synthesis reagents (Bio-Rad). qRT-PCR was carried out with SYBR Select Master Mix (Thermo Fisher Scientific) on a Bio-Rad CFX96 thermocycler. Primers used in this study are listed in Table S10.

Data plotting and statistical analyses were carried out with GraphPad Prism software (version 9.1.2). Sample sizes and statistical methods used are indicated in figure and table legends. Unless otherwise stated, individual data points are plotted as mean±s.d. Paralysis assays are plotted as Kaplan–Meier survival curves with data aggregated from at least three independent experiments.

We thank Drs Harold Smith and Ilhan Akan of the National Institute of Diabetes and Digestive and Kidney Diseases Genomic Core Laboratory for use of their resources, reagents and expertise. We also thank the Caenorhabditis Genetics Center, which is funded by National Institutes of Health Office of Research Infrastructure Programs (P40 OD010440), for supplying published strains. We thank Dr Peter Askjaer (Andalusian Center for Developmental Biology, Universidad Pablo de Olavide) for generation of the FLP driver strains and expert advice on experimental design with FLP/FRT systems. Finally, we thank Dr William Mair (Department of Stem Cell and Regenerative Biology, Harvard University) for generation of the SKI LODGE system and for providing the previously unpublished WBM1456 strain.