Slo2 potassium channel function depends on RNA editing-regulated expression of a SCYL1 protein

Slo2 potassium channels play important roles in neuronal function, and their mutations in humans may cause epilepsies and cognitive defects. However, it is largely unknown how Slo2 is regulated by other proteins. Here we show that the function of C. elegans Slo2 (SLO-2) depends on adr-1, a gene important to RNA editing. ADR-1 promotes SLO-2 function not by editing the transcripts of slo-2 but those of scyl-1, which encodes an orthologue of mammalian SCYL1. Transcripts of scyl-1 are greatly decreased in adr-1 mutants due to deficient RNA editing at a single adenosine in their 3’-UTR. SCYL-1 physically interacts with SLO-2 in neurons. Single-channel open probability (Po) of neuronal SLO-2 is ~50% lower in scyl-1 knockout mutant than wild type. Moreover, human Slo2.2/Slack Po is doubled by SCYL1 in a heterologous expression system. These results suggest that SCYL-1/SCYL1 is an evolutionarily conserved regulator of Slo2 channels.

RNA editing is an evolutionally conserved post-transcriptional process catalyzed by ADARs (adenosine deaminases acting on RNA) (Gott and Emeson, 2000;Jin et al., 2009). ADARs convert adenosine (A) to inosine (I) in double-stranded RNA. Since inosine is interpreted as guanosine (G) by cellular machineries (Basilio et al., 1962), A-to-I RNA editing may alter the function of a protein by changing its coding potential, or regulate gene expression through altering alternative splicing, microRNA processing, or RNA interference (Deffit and Hundley, 2016;Nishikura, 2016). Human and mouse each has three ADARs: ADAR1, ADAR2 and ADAR3 (Chen et al., 2000;Kim et al., 1994;Melcher et al., 1996). ADAR1 and ADAR2 possess deaminase activity and catalyze the A-to-I conversion (Tan et al., 2017), whereas ADAR3 is catalytically inactive with regulatory roles in RNA editing (Nishikura, 2016). Millions of A-to-I editing sites have been detected in the human transcriptome through RNA-seq, with the vast majority of them found in non-coding regions (Nishikura, 2016). Biological effects of RNA editing at coding regions have been revealed for a variety of genes, including those encoding ligand-and voltage-gated ion channels and G protein-coupled receptors (Bhalla et al., 2004;Brusa et al., 1995;Burns et al., 1997;Gonzalez et al., 2011;Huang et al., 2012;Lomeli et al., 1994;Palladino et al., 2000;Rula et al., 2008;Sommer et al., 1991;Streit et al., 2011). However, little is known about the roles of RNA editing in non-coding regions (Nishikura, 2016).
In a genetic screen for suppressors of a sluggish phenotype caused by expressing a hyperactive SLO-2 in worms, we isolated mutants of several genes, including adr-1, which encodes one of two ADARs in C. elegans (ADR-1 and ADR-2). While ADR-2 has deaminase activity and plays an indispensable role in the A-to-I conversion, ADR-1 is catalytically inactive but can promote RNA editing by binding to selected target mRNA and tethering ADR-2 to RNA substrates (Ganem et al., 2019;Rajendren et al., 2018;Washburn et al., 2014). We found that loss-of-function (lf) mutations of adr-1 inhibit SLO-2 function through impairing RNA editing of scyl-1, which encodes an orthologue of human and mouse SCYL1. In adr-1(lf) mutants, a lack of A-to-I conversion at a specific site in scyl-1 3'-UTR causes reduced scyl-1 expression. Knockout of scyl-1 severely reduces SLO-2 current in worms whereas coexpression of SCYL1 with human Slack in Xenopus oocytes greatly augments channel activity. These results suggest that SCYL-1/SCYL1 proteins likely play an evolutionarily conserved role in physiological functions of Slo2 channels. Mutations or knockout mammalian SCYL1 may cause neural degeneration, intellectual disabilities, and liver failure, but the underlying mechanisms are unclear (Lenz et al., 2018;Li et al., 2019;Shohet et al., 2019;Spagnoli et al., 2019). The revelation of SCYL-1/SCYL1 as a protein important to Slo2 channels suggests a potential link between diseases caused by SCLY1 mutations and Slo2 channel functions.

Results
adr-1 mutants suppress sluggish phenotype of slo-2(gf) In a genetic screen for mutants that suppressed a sluggish phenotype caused by an engineered hyperactive or gain-of-function (gf) SLO-2 (Liu et al., 2018), we isolated two mutants (zw80 and zw81) of the adr-1 gene, as revealed by analyses of whole-genome sequencing data. zw80 and zw81 carry nonsense mutations leading to premature stops at tryptophan (W) 366 and W33, respectively ( Figure 1A). slo-2(gf) worms showed greatly decreased locomotion speed compared with wild type, and this phenotype was substantially alleviated in slo-2(gf);adr-1(lf) double mutants ( Figure 1B). To confirm that the suppression of slo-2(gf) phenotype resulted from mutations of adr-1 rather than that of another gene, we created a new adr-1 mutant allele (zw96) by introducing a premature stop codon at serine (S) 333 ( Figure 1A) using the CRISPR/Cas9 approach. The sluggish phenotype of slo-2(gf) was similarly suppressed by adr-1(zw96), which, by itself, did not enhance locomotion speed ( Figure 1B). Expression of wild-type adr-1 under the control of the pan-neuronal rab-3 promotor (Prab-3) in slo-2(gf);adr-1(zw96) unmasked the sluggish phenotype ( Figure 1B). These results indicate that the sluggish phenotype of slo-2(gf) is mainly caused by SLO-2 hyperactivity in neurons, and that neuronal function of SLO-2(gf) depends on ADR-1.
We suspected that the suppression of SLO-2(gf) by adr-1(lf) resulted from deficient RNA-editing. If so, adr-2(lf) might also suppress the sluggish phenotype of slo-2(gf) because ADR-2 is required for RNA editing. Indeed, the sluggish phenotype of slo-2(gf) worms was substantially alleviated in slo-2 (gf);adr-2(lf) double mutants (Figure 2A), and the augmenting effect of slo-2(gf) on VA5 whole-cell outward currents was mostly eliminated by adr-2(lf) ( Figure 2B). Furthermore, adr-2(lf) brought VA5 whole-cell currents below the wild-type level ( Figure 2B), which presumably resulted from reduced activities of wild-type SLO-2. These results suggest that RNA editing is important to SLO-2 function in neurons. ADR-1 has two doublestranded RNA-binding motifs (dsRBM) and a pseudodeaminase domain. (B) Mutations of adr-1 mitigated an inhibitory effect of hyperactive or gain-of-function (gf) SLO-2 on locomotion through acting in neurons. adr-1 rescue was achieved by expressing GFP-tagged wild-type ADR-1 in neurons under the control of Prab-3 (same in C and D). Sample sizes were 10-12 in each group. (C) adr-1(zw96) reduced an augmenting effect of slo-2(gf) on motor neuron whole-cell outward currents. Pipette solution I and bath solution I were used. Sample sizes were 7 wild type, 8 slo-2(gf), 9 slo-2(gf);adr-1(zw96), and 8 slo-2(gf);adr-1(zw96) rescue. (D) adr-1(zw96) mitigated an inhibitory effect of slo-2(gf) on postsynaptic current (PSC) bursts at the neuromuscular junction. The vertical dotted lines over the sample traces mark PSC bursts, which are defined as an apparent increase in PSC frequency accompanied by a sustained current (downward baseline shift) lasting >3 s. Pipette solution II and bath solution I were used. Sample sizes were 12 wild type, and 7 in each of the remaining groups. All values are shown as mean ± SE. The asterisks indicate statistically significant differences between indicated groups (*p<0.05, ***p<0.001) based on either two-way (C) or one-way (D) ANOVA with Tukey's post hoc tests. The online version of this article includes the following source data for figure 1: Source data 1. Raw data and numerical values for data plotted in Figure 1.

ADR-1 is expressed in neurons and localized in the nucleus
The expression pattern of adr-1 was examined by expressing GFP under the control of adr-1 promoter (Padr-1). In transgenic worms, strong GFP expression was observed in the nervous system, including ventral cord motor neurons and many neurons in the head and tail, while weak GFP expression was observed in the intestine and body-wall muscles ( Figure 3A). We then examined the subcellular localization pattern of ADR-1 by expressing GFP-tagged full-length ADR-1 (ADR-1::GFP) under the control of Prab-3. We found that ADR-1::GFP is localized in the nucleus, as indicated by its colocalization with the mStrawberry-tagged nucleus marker HIS-58 (Liu et al., 2018) in ventral cord motor neurons ( Figure 3B). Loss-of-function mutation of adr-2 suppressed the effects of gain-of-function (gf) slo-2 on locomotion and motor neuron whole-cell currents. (A) adr-2(gv42) alleviated an inhibitory effect of slo-2(gf) on locomotion speed. The sample size was 10-12 in each group. (B) adr-2(gv42) largely reversed an augmenting effect of slo-2(gf) on whole-cell currents in VA5 motor neuron. Sample sizes were 11 wild type, 8 slo-2(gf), 11 slo-2(gf);adr-2(gv42), and 10 adr-2(gv42). All data are shown as mean ± SE. Pipette solution I and bath solution I were used. The asterisks indicate statistically significant differences (*p<0.05; ***p<0.001) whereas 'ns' stands for 'no significant difference' between the indicated groups based on either one-way (A) or two-way (B) ANOVA with Tukey's post hoc tests. The online version of this article includes the following source data for figure 2: Source data 1. Raw data and numerical values for data plotted in Figure 2.
To determine whether adr-1 is co-expressed with slo-2, we crossed the Padr-1::GFP transgene into an existing strain expressing Pslo-2::mStrawberry (Liu et al., 2018). We found that the expression patterns of adr-1 and slo-2 overlapped extensively in the nervous system ( Figure 3C). For example, the majority of ventral cord motor neurons and numerous head neurons were colabeled by GFP and mStrawberry ( Figure 3C). The occasional non-overlapping expressions of GFP and mStrawberry in ventral cord motor neurons probably resulted from mosaic expression of the transgenes.
ADR-1 regulates neurotransmitter release through SLO-2 SLO-2 is the primary conductor of delayed outward currents in C. elegans cholinergic motor neurons (Liu et al., 2014). We wondered whether the function of native SLO-2 channels in motor neurons depends on ADR-1. Consistent with our previous report (Liu et al., 2014), VA5 delayed outward currents were dramatically smaller and VA5 resting membrane potential was much less hyperpolarized in slo-2(lf) than wild type. While adr-1(lf) also caused significantly decreased outward currents and less hyperpolarized resting membrane potential in VA5, it did not produce additive effects when combined with slo-2(lf) ( Figure 4A-C). These results suggest that adr-1(lf) affects motor neuron outward currents and resting membrane potential through SLO-2.  Sample sizes were 8 wild type, 7 slo-2(lf), 9 adr-1(zw96), 7 slo-2(lf);adr-1(zw96), and 9 adr-1 (zw96) rescue. (C) Resting membrane potentials of VA5. Sample sizes were 6 wild type, and 7 in each of the remaining groups. (D) Representative traces of spontaneous PSCs with PSC bursts marked by vertical dotted lines. (E) Comparisons of PSC burst properties. Sample sizes were 8 slo-2(lf);adr-1(zw96), 6 adr-1(zw96) rescue, and 12 in each of the remaining groups. All values are shown as mean ± SE. The asterisks indicate statistically significant differences (*p<0.05, ***p<0.001) compared with wild type whereas 'ns' stands for no significant difference between the indicated groups based on either two-way (B) or one-way (C and E) ANOVA with Tukey's post hoc tests. Pipette solution I and bath solution I were used in (A) and (C). Pipette solution II and bath solution I were used in (D).

Figure 4 continued on next page
We next determined whether adr-1(lf) also alters PSC bursts. We found that adr-1(lf) caused an increase in the duration and mean charge transfer rate of PSC bursts without altering the burst frequency compared with wild type ( Figure 4D and E). These phenotypes of adr-1(lf) were similar to those of slo-2(lf) and did not become more severe in the double mutants ( Figure 4D and E), suggesting that ADR-1 modulates neurotransmitter release through SLO-2. The similar effects of adr-1 (lf) and slo-2(lf) on PSC bursts are in contrast to their differential effects on VA5 outward currents and resting membrane potential. This discrepancy suggests that reducing SLO-2 activity beyond a certain threshold may produce similar effects on PSC bursts as does slo-2(lf).

ADR-1 regulates SLO-2 function through SCYL-1
Given that our results suggest that RNA editing is important to SLO-2 function, we determined whether adr-1(lf) causes deficient editing or decreased expression of slo-2 mRNA by comparing RNA-seq data between adr-1(lf) and wild type. The adr-1(zw96) allele was chosen for these analyses to avoid complications by potential mutations of other genes introduced into the genome during the generation of the other adr-1 mutants (zw80 and zw81). Unexpectedly, no RNA editing event was detected in slo-2 transcripts, and slo-2 mRNA level was similar between wild type and the adr-1 mutant (Figure 4-figure supplement 1). These results suggest that ADR-1 might regulate SLO-2 function through RNA editing of another gene.
A previous study identified 270 high-confidence editing sites in transcripts of 51 genes expressed in C. elegans neurons (Washburn et al., 2014). We suspected that the putative molecule mediating the effect of ADR-1 on SLO-2 is encoded by one of these genes, and the mRNA level of this gene is reduced in adr-1(lf). Therefore, we compared transcript expression levels of these genes (excluding those encoding transposons) in our RNA-Seq data between wild type and adr-1(zw96). Most of these genes showed either no decrease or only a small decrease in expression, but two of these genes, rncs-1 and scyl-1, were reduced greatly in adr-1(lf) compared with wild type ( Figure 5). rncs-1 is not a conceivable candidate for the putative SLO-2 regulator because it is a non-coding gene expressed in the hypodermis and vulva (Hellwig and Bass, 2008). We therefore focused our analyses on scyl-1, which encodes an orthologue of mammalian SCYL1 important to neuronal function and survival (Pelletier, 2016). Like its mammalian homologs, SCYL-1 has an amino-terminal kinase domain that lacks residues critical to kinase activity, and a central domain containing five HEAT repeats (HEAT for Huntingtin, elongation factor 3, protein phosphatase 2A, yeast kinase TOR1) (Pelletier, 2016). SCYL-1 shares 38% identity and 60% similarity with human SCYL1. Notably, amino acid sequence in the HEAT domain, which is often highly degenerative (Pelletier, 2016), shows a very high level of sequence homology (53% identity and 76% similarity) between these two proteins (Figure 5-figure supplement 1).
We examined the expression pattern of scyl-1 by expressing GFP under the control of scyl-1 promoter (Pscyl-1). Because another gene (lap-2) resides~2 kb upstream of scyl-1 (www.wormbase.org), we first expressed GFP under the control of 2 kb Pscyl-1. However, no GFP signal was detected in transgenic worms (not shown). We then used an in vivo homologous recombination approach to express a Pscyl-1::GFP transcriptional fusion that included a much longer sequence upstream of the scyl-1 initiation site. Specifically, a 0.5 kb genomic fragment upstream of the scyl-1 initiation site was fused to GFP in a plasmid, which was co-injected with a fosmid covering part of the scyl-1 coding region and 32 kb sequence upstream of the initiation site into wild type worms. In vivo homologous recombination between the plasmid and the fosmid is expected to result in a promoter::GFP transcriptional fusion that includes all the upstream genomic sequence in the fosmid. Transgenic worms from coinjection of the plasmid and the fosmid showed GFP signal in a variety of cells ( Figure 6), suggesting distant upstream sequences are required for scyl-1 expression. To determine how the expression pattern of scyl-1 correlates with that of slo-2, we crossed the transgene into the Pslo-2:: mStrawberry strain, and examined the expression patterns of GFP and mStrawberry. Co-expression of scyl-1 and slo-2 was observed in many ventral cord motor neurons ( Figure 6). However, most other neurons expressing slo-2 (e. g. head and tail neurons) did not appear to express scyl-1. In addition, scyl-1 expression was detected in some cells that did not express slo-2, including the excretory cell, spermatheca, uterine ventral cells, and intestinal cells ( Figure 6).
We next determined whether SCYL-1 is important to SLO-2 function. To this end, we created a mutant, scyl-1(zw99), by introducing a stop codon after isoleucine 152 using the CRISPR/Cas9 approach, and examined the effect of this mutation on VA5 delayed outward currents. scyl-1(zw99) showed a substantial decrease in VA5 outward currents compared with wild type; this phenotype could be rescued by expressing wild-type SCYL-1 in neurons, and was non-additive with that of either slo-2(lf) or adr-1(zw96) ( Figure 7A). These results suggest that SCYL-1 and ADR-1 likely act in a common pathway to contribute to SLO-2 function.
The decrease of delayed outward currents in scyl-1(lf) could have resulted from either reduced expression or reduced function of SLO-2. We first determined whether scyl-1(lf) alters SLO-2 expression by crossing a stable (near 100% penetrance) Prab-3::SLO-2::GFP transgene from an existing transgenic strain of wild-type genetic background (Liu et al., 2018) into scyl-1(zw99), and comparing GFP signal between the two strains. We found that GFP signal in the ventral nerve cord was similar between wild type and the scyl-1 mutant ( Figure 7B), suggesting that SCYL-1 does not regulate SLO-2 expression. We then determined whether SCYL-1 regulates SLO-2 function by obtaining inside-out patches from VA5 and analyzing SLO-2 single-channel properties. SLO-2 showed >50% decrease in open probability (P o ) without a change of single-channel conductance in scyl-1(zw99) compared with wild type, and this mutant phenotype was completely rescued by neuronal expression of wild-type SCYL-1 ( Figure 8A). Analyses of single-channel open and closed events revealed that SLO-2 has at least two open states and three closed states, and that the decreased P o of SLO-2 in scyl-1(lf) mainly resulted from shorter and fewer long openings ( Figure 8B) and a larger proportion of the longest closings ( Figure 8C).
The observed effects of scyl-1(lf) on SLO-2 single-channel properties suggest that SCYL-1 may physically interacts with SLO-2. To address this possibility, we performed bimolecular fluorescence complementation (BiFC) (Hu et al., 2002) and coimmunoprecipitation assays, which indicate whether these two proteins are physically very close in vivo and whether they coexist in a molecular complex, Figure 5. Normalized transcript expression levels of selected genes in adr-1(zw96) mutant. The genes were selected based on the detection of ADR-1dependent RNA editing events in their transcripts reported in an earlier study (Washburn et al., 2014). Transcript expression level of each gene in the mutant is normalized by that in the wild type. Shown are mean ± SE from three biological replicates of RNA-seq experiments. The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Raw data and numerical values for data plotted in Figure 5. respectively. In both assays, we determined whether full-length SCYL-1 interacts with either the fulllength, the amino-terminal portion (amino acids 1-317), or the carboxyl-terminal portion (amino acids 318-1107) of SLO-2 ( Figure 9A). In the BiFC assay, SCYL-1 and SLO-2 were fused to the Nand C-terminal portions of YFP (YFPa and YFPc), respectively ( Figure 9A). A detection of YFP signal would indicate physical closeness of the two fusion proteins. We observed YFP fluorescence in ventral cord motor neurons when either the full-length or the C-terminal of SLO-2 was used but not when the N-terminal was used in the assays ( Figure 9B). The coimmunoprecipitation assay was performed with homogenates of worms expressing HA-tagged SCYL-1 and GFP-tagged SLO-2. We found that the SCYL-1 immunoprecipitated with either full-length SLO-2 or SLO-2 C-terminal but not SLO-2 N-terminal ( Figure 9C), which is in agreement with the BiFC results. Thus, both the BiFC and coimmunoprecipitation results suggest that SCYL-1 physically interacts with SLO-2, and this interaction is mediated by SLO-2 carboxyl terminal.

scyl-1 expression depends on RNA editing at a specific 3'-UTR site
Our RNA-Seq data revealed eight high-frequency (>15%) adenosine-to-guanosine editing sites in scyl-1 transcripts of wild type ( Figure 10A). All these editing sites are located within a predicted 746 bp hair-pin structure in the 3' end of scyl-1 pre-mRNA, which contains an inverted repeat with >98% complementary base pairing ( Figure 10B). Interestingly, RNA editing at only one of these sites was significantly deficient (by 74%) in adr-1(zw96) compared with wild type ( Figure 10A). Sanger sequencing of scyl-1 mRNA and the corresponding genomic DNA from wild type, adr-1(zw96), and adr-2(gv42) showed that RNA editing at this site was deficient in both the adr-1 and adr-2 mutants whereas that at an adjacent site was deficient only in the adr-2 mutant ( Figure 10C), suggesting that RNA editing at the site impaired by adr-1(lf) might be important to scyl-1 expression. To address this possibility, we created transgenic worms expressing a Prab-3::GFP transcriptional fusion (wp1923), in which a genomic DNA fragment covering part of the last exon of scyl-1 and 5 kb downstream sequence was fused in-frame to the 3'-end of GFP coding sequence ( Figure 10D). We also created transgenic worms expressing a modified plasmid (wp1924), in which adenosine (A) at the specific ADR-1-dependent editing site was changed to guanosine (G) to mimic the edited nucleotide ( Figure 10D). GFP signal was observed in worms harboring wp1924 but no GFP signal was detected in worms harboring wp1923 ( Figure 10E). While observation of GFP signal in the wp1924 strain was expected, the complete absence of GFP signal in the wp1923 strain caught us by surprise. To better understand the role of ADR-1 in scyl-1 mRNA expression, we integrated the wp1924 transgene into the wild-type genome, crossed it into adr-1(zw96), and compared GFP signal in ventral cord motor neurons between wild type and the mutant. GFP signal was~50% weaker in the mutant than wild type ( Figure 10F and G). Because the transgene mimicked the edited 3'-UTR sequence of scyl-1 mRNA, the weaker GFP signal in the mutant than wild type suggests that ADR-1 can also increase scyl-1 mRNA level through a post-editing effect.
If ADR-1 does enhance scyl-1 mRNA level through a post-editing mechanism in addition to RNA editing, it likely perform this function through interacting with some other proteins, and such proteins might be identified by screening for mutants showing decreased GFP signal from the wp1924 transgene. Indeed, eleven mutants with reduced GFP expression were isolated from~5000 mutagenized haploid genomes in a pilot genetic screen, as shown by two examples (Figure 10-figure  supplement 1). The deficiency of GFP expression in these mutants was related to the function of scyl-1 3'-UTR because GFP expression from a control transgene with unc-10 3'-UTR was not Figure 7. SCYL-1 contributes to motor neuron outward currents through SLO-2. (A) Sample whole-cell current traces of VA5 motor neurons and the current-voltage relationships. Sample sizes were 9 wild type, 7 slo-2(lf), 10 scyl-1(zw99), 7 slo-2(lf);scyl-1(zw99), 7 adr-1(zw96);scyl-1(zw99), and 7 scyl-1(zw99) rescue. The rescue strain was created by expressing wild-type scyl-1 under the control of Prab-3. All values are shown as mean ± SE. The asterisks (***) and pound signs ( ### ) indicate statistically significant differences (p<0.001) between the indicated groups and from wild type, respectively, whereas 'ns' stands for no significant difference between the indicated groups (two-way ANOVA with Tukey's post hoc tests). (B) GFP signal in ventral cord motor neurons was indistinguishable between wild type and scyl-1(zw99) worms expressing GFP-tagged full-length SLO-2 under the control of Prab-3. Scale bar = 20 mm. The online version of this article includes the following source data for figure 7: Source data 1. Raw data and numerical values for data plotted in Figure 7. compromised in these mutants ( Figure 10-figure supplement 1). These results favor the notion that ADR-1 may also interact with other molecules after RNA editing to promote scyl-1 expression.

Human Slo2.2/Slack is regulated by SCYL1
The HEAT domain of SCYL proteins is important to protein-protein interactions but generally varies considerably in amino acid sequence for interactions with specific protein partners (Yoshimura and Hirano, 2016). The high level of sequence homology of the HEAT domain between mammalian SCYL1 and worm SCYL-1 ( Figure 5-figure supplement 1) promoted us to test whether mammalian Slo2.2/Slack is also regulated by SCYL1. We expressed human Slo2.2 (hSlo2.2) either alone or together with mouse SCYL1 in Xenopus oocytes, and analyzed hSlo2.2 single-channel properties in inside-out patches. SCYL1 increased hSlo2.2 P o greatly without altering the single-channel conductance ( Figure 11A). The channel has at least two open states and three closed states ( Figure 11B-D). While events of both the open states and the two shorter closed states were numerous, those of the longest closed state were infrequent. Nevertheless, the longest closed state had a major impact on P o because of its rather long duration. Dwell time analyses indicate that SCYL1 increased hSlo2.2  Figure 11B-D). The overall effect of SCYL1 on hSlo2.2 P o is similar to that of SCYL-1 on SLO-2 P o in wild-type worms (Figure 8). Taken together, these results suggest that a physiological function of both mammalian SCYL1 and worm SCYL-1 is to enhance Slo2 channel activities.

Discussion
This study shows that both ADR-1 and SCYL-1 are critical to SLO-2 physiological function in neurons. While ADR-1 enhances SLO-2 function indirectly through promoting SCYL-1 expression, SCYL-1 regulates SLO-2 through physical interactions. These conclusions are supported by multiple lines of evidence, including the isolation of adr-1(lf) mutants as suppressors of SLO-2(gf), the inhibition of SLO-2 activities by either adr-1(lf) or scyl-1(lf), the reduction of scyl-1 transcript expression in adr-1(lf), the correlation between scyl-1 mRNA level and RNA editing at its 3'-UTR, the evidence of physical interactions between SCYL-1 and SLO-2, and the reduction of SLO-2 P o due to changes in the dwell times of open and closed events in scyl-1(lf). Importantly, we found that the human Slo2.2/Slack is also regulated by SCYL1.
The biological significance of RNA editing at non-coding regions is only beginning to be appreciated. A recent study with C. elegans identified many neuron-specific A-to-I editing sites in the 3'-UTR of clec-41, and showed that elimination of these editing events by adr-2 knockout compromises clec-41 expression and impairs a chemotaxis behavior (Deffit et al., 2017). However, it remains to be determined how clec-41 expression is controlled by these editing events, and whether the chemotaxis defect of adr-2(lf) mutant is caused by decreased clec-41 expression. In the present study, we demonstrate that an A-to-I RNA editing event at the 3'UTR of scyl-1 controls its expression, and that SCYL-1 contributes to neuronal whole-cell currents through a direct effect on SLO-2. The results of these two studies have provided a glimpse of the biological roles of 3'-UTR RNA editing in gene expression and neuronal function.   -1(zw96), and adr-2(gv42), and of the corresponding wild type genomic DNA. Two editing sites in wild type mRNA (indicated by arrows) display a mixture of green (adenosine) and black (guanosine) peaks. While both editing events are non-existent in adr-2(gv42), only one of them is inhibited by adr-1(zw96). (D) Diagram of two GFP reporter constructs (wp1923 and wp1924) used to confirm the role of the ADR-1-dependent editing site in gene expression. GFP was placed under the control of Figure 10 continued on next page Our results demonstrate that RNA editing at a single site in the 3'-UTR could have a profound effect on gene expression. The A-to-I conversion at the ADR-1-regulated editing site increases base pairing in the putative double-stranded structure of scyl-1 3'-UTR ( Figure 10B). Increased base paring in a double-stranded RNA generally facilitates RNA degradation. It is therefore intriguing how such an editing event may cause increased gene expression. One

Figure 10 continued
Prab-3 and fused to the last exon (blue) of scyl-1 followed by 5 kb downstream genomic sequence. The red bars indicate the inverted repeat sequences that form the double-stranded RNA in the hair-pin structure (B). wp1923 contains the intact genomic sequence of scyl-1 3'-UTR, whereas wp1924 differs from it in an A-to-G conversion mimicking the ADR-1-dependent editing. (E) Effects of the A-to-G conversion on GFP reporter expression. Shown are fluorescent and corresponding DIC images of transgenic worms harboring either wp1923 or wp1924. GFP expression in the head and ventral nerve cord (VNC) was observed only in worms harboring wp1924. The diffused signal below the VNC in fluorescent images was from auto-fluorescence of the intestine (Int). Scale bar = 20 mm. (F) GFP expression from wp1924 was decreased in adr-1(zw96) compared with that in wild type. Scale bar = 20 mm. (G) Statistical comparison of GFP intensity in the VNC between wild type (n = 16) and adr-1(zw96) (n = 19). (***p<0.001, unpaired t-test). The online version of this article includes the following source data and figure supplement(s) for figure 10: Source data 1. Raw data and numerical values for data plotted in Figure 10.   Figure 8 legend. (D) Dwell time histograms and statistical comparisons of the long closed events. Closed events that were >30 ms in duration of all recordings in each group were pooled together to construct the dwell time histogram. The average duration and frequency of these events were compared between the two groups. Each dot represents the mean value of one recording. Sample sizes were 13 in both groups. All values are shown as mean ± SE. The asterisks indicate a significant difference compared between the indicated groups (*p<0.05, ***p<0.001, unpaired t-test). The online version of this article includes the following source data for figure 11: Source data 1. Raw data and numerical values for data plotted in Figure 11. possibility is that editing at this site helps maintain mRNA stability through recruiting other regulatory proteins to the 3'-UTR. The isolation of mutants exhibiting diminished GFP expression from a reporter construct containing the edited scyl-1 3'-UTR but not the unc-10 3'-UTR (Figure 10-figure supplement 1) indicates potential existence of such regulatory proteins. In combination with the observation that GFP signal was undetectable in transgenic worms expressing a reporter construct containing the non-edited scyl-1 3'-UTR, this result suggests that RNA editing at the ADR-1-dependent site in scyl-1 3'-UTR is likely important to recruiting the putative regulatory proteins to increase mRNA stability.
scyl-1 has four different transcripts with identical 5'-UTR and coding sequence but different 3'-UTRs of variable lengths (ranging from 167 to 1862 nucleotides) and sequences (www.wormbase.org). The ADR-1-dependent editing site exists only in the transcript with the longest 3'-UTR. The presence of an ADR-1-regulated editing site in this but not the other transcripts suggests that ADR-1 may regulate scyl-1 expression in a cell-specific manner depending on where the specific transcript is expressed. Interestingly, the 3'-UTR of human SCYL1 transcripts (NM_020680.4) also has a high probability of forming hair-pin structures based on software prediction (https://rna.urmc.rochester.edu/RNAstructureWeb/Servers). It remains to be determined whether human SCYL1 transcripts are also edited in the 3'-UTR, and if so, whether the editing regulates their expression. SCYL1 proteins are evolutionarily conserved proteins that share an N-terminal pseudokinase domain (Manning et al., 2002;Pelletier, 2016). Results of previous studies with cultured cells suggest that SCYL1 may regulate intracellular trafficking between the Golgi apparatus and the ER (Burman et al., 2008;Burman et al., 2010), and facilitate nuclear tRNA export by acting at the nuclear pore complex (Chafe and Mangroo, 2010). Mutations of SCYL1 in humans are associated with a variety of disorders, including neurodegeneration, intellectual disabilities, and liver failure (Lenz et al., 2018;Li et al., 2019;Schmidt et al., 2015;Shohet et al., 2019;Spagnoli et al., 2019). Mice with SCYL1 deficiency develop an early onset and progressive neurodegenerative disorder (Pelletier et al., 2012). However, it is unclear whether the documented mutant phenotypes of SCYL1 are related to its roles in intracellular trafficking and nuclear tRNA export (Pelletier, 2016). This study brings into view a new potential mechanism for SCYL1 mutation-associated disorders: impairing Slo2 channel function. What might be the molecular mechanism through which SCYL-1 enhances SLO-2 activity? Since SCYL-1 physically associates with SLO-2, and enhances SLO-2 singlechannel P o by altering the open and closed states, it likely regulates channel function either directly or through a closely associated protein. Although the exact mechanism remains to be determined, the similar effects of SCYL-1 and SCYL1 on Slo2 channels suggest that they likely play an important role in Slo2 physiological functions across animal species.
The expression patterns of scyl-1 and slo-2 largely do not overlap. Although they are coexpressed in ventral cord motor neurons, most other neurons expressing slo-2 do not express scyl-1, suggesting that the regulatory effect of SCYL-1 on SLO-2 is cell-and tissue-specific. The expression of scyl-1 in cells not expressing slo-2 suggests that SCYL-1 physiological functions are not limited to regulating SLO-2. In mouse, SCYL1 and Slo2.2 are both expressed in the hippocampus and cerebellum but their expression patterns do not completely overlap (Joiner et al., 1998;Schmidt et al., 2007). Conceivably, the regulation of hSlo2.2 by SCYL1 might also occur in a cell-and tissue-specific manner, and SCYL1 likely performs other physiological functions. The latter possibility is supported by the pleiotropic phenotypes observed in patients and mice with SCYL1 mutations (Lenz et al., 2018;Li et al., 2019;Pelletier et al., 2012;Schmidt et al., 2015;Shohet et al., 2019;Spagnoli et al., 2019).
In summary, this study demonstrates that ADAR-mediated RNA editing controls the expression of SCYL-1, which interacts with SLO-2 to allow SLO-2 perform its physiological functions. Moreover, this study shows that this regulatory mechanism is conserved with mammalian SCYL1 and Slo2. Our findings reveal a new molecular mechanism of Slo2 channel regulation, and provide the bases for investigating how physiological functions of human Slo2 are regulated by SCYL1, and whether the neurodegeneration and intellectual disability phenotypes of SCYL1 mutations are related to Slo2 channel dysfunction. Continued on next page contains 32 kb of scyl-1 upstream sequence and part of its coding region, into the lin-15(n765) strain along with a lin-15 rescue plasmid to serve as a transformation marker. To assay the effect of the identified adenosine site at the 3'UTR of scyl-1 on gene expression, a 5.1 kb genomic DNA fragment covering part of the scyl-1 last exon and subsequent sequence was cloned and fused in-frame to GFP using the primers 5'-AATGCTAGCATGCAGGCTAGAAATGAAGCTCG (forward) and 5'-TA TGGGCCCGAAATCAGCATCTTTGACGAA (reverse). To mimic the A-to-I editing at the identified specific site, a second plasmid was made by mutating the specific adenosine to guanosine in the above plasmid. The two resultant plasmids were injected into lin-15(n765), respectively, with a lin-15 rescue plasmid as the transgenic marker. Images of transgenic worms were taken with a digital CMOS camera (Hamamatsu, C11440-22CU) mounted on a Nikon TE2000-U inverted microscope equipped with EGFP/FITC and mCherry/Texas Red filter sets (49002 and 49008, Chroma Technology Corporation, Rockingham, VT, USA).

Behavioral assay
Locomotion velocity was determined using an automated locomotion tracking system as described previously . Briefly, a single adult hermaphrodite was transferred to an NGM plate without food. After allowing~30 s for recovery from the transfer, snapshots of the worm were taken at 15 frames per second for 30 s using a IMAGINGSOURCE camera (DMK37BUX273) mounted on a stereomicroscope (LEICA M165FC). The worm was constantly kept in the center of the view field with a motorized microscope stage (OptiScanTM ES111, Prior Scientific, Inc, Rockland, MA, USA). Both the camera and the motorized stage were controlled by a custom program (Source code 1) running in MATLAB (The MathWorks, Inc, Natick, MA).

RNA-seq and data analysis
Total RNA was extracted from young adult-stage worms using TRIzol Reagent (Invitrogen) and treated with TURBO DNase (Ambion). RNA-seq was performed by Novogene Corp. Sacramento, CA. Raw reads ware filtered using Trim Galore software (http://www.bioinformatics.babraham.ac.uk/ projects/trim_galore/) to remove reads containing adapters or reads of low quality. The filtered reads were mapped to C. elegans genome (ce11) using TopHat2 (Kim et al., 2013). The gene expression level is estimated by counting the reads that map to exons.

C. elegans electrophysiology
Adult hermaphrodites were used in all electrophysiological experiments. Worms were immobilized and dissected as described previously (Liu et al., 2007). Borosilicate glass pipettes were used as electrodes for recording whole-cell currents. Pipette tip resistance for recording muscle cell currents was 3-5 MW whereas that for recording motor neuron currents was~20 MW. The dissected worm preparation was treated briefly with collagenase and perfused with the extracellular solution for 5 to 10-fold of bath volume. Classical whole-cell configuration was obtained by applying a negative pressure to the recording pipette. Current-and voltage-clamp experiments were performed with a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA) and the Clampex software (version 10, Molecular Devices). Data were sampled at a rate of 10 kHz after filtering at 2 kHz. Spontaneous membrane potential changes were recorded using the current-clamp technique without current injection. Motor neuron whole-cell outward currents were recorded by applying a series of voltage steps (À60 to +70 mV at 10 mV intervals, 600 ms pulse duration) from a holding potential of À60 mV. Spontaneous PSCs were recorded from body-wall muscle cells at a holding potential of À60 mV. Two bath solutions and three pipette solutions were used in electrophysiological experiments as specified in figure legends. Bath solution I contained (in mM) 140 NaCl, 5 KCl, 5 CaCl 2 , 5 MgCl 2 , 11 dextrose and 5 HEPES (pH 7.2). Bath solution II contained (in mM) 100 K + gluconate, 50 KCl, 1 Mg 2+ gluconate, 0.1 Ca 2+ gluconate and 10 HEPES (pH 7.2). Pipette solution I contained (in mM) 120 KCl, 20 KOH, 5 Tris, 0.25 CaCl 2 , 4 MgCl 2 , 36 sucrose, 5 EGTA, and 4 Na 2 ATP (pH 7.2). Pipette solution II differed from pipette solution I in that 120 KCl was substituted by K + gluconate. Pipette solution III contained (in mM) 150 K + gluconate, 1 Mg 2+ gluconate and 10 HEPES (pH 7.2).

Data analyses for electrophysiology
Amplitudes of whole-cell currents in response to voltage steps were determined from the mean current during the last 100 ms of the 1200 ms voltage pulses using the Clampfit software. The duration and charge transfer of PSC bursts were quantified with Clampfit software (version 10, Molecular Devices) as previously described (Liu et al., 2013). The frequency of PSC bursts was counted manually. For single channel analysis, the QuB software (https://qub.mandelics.com/) was used to fit open and closed times to exponentials, and to quantify the t values and relative areas of the fitted components, which were automatically determined by the software. Single-channel conductance was calculated by dividing the single-channel current amplitude (determined from a Gaussian fit to the amplitude histogram) by the holding voltage. The first 30 s recording of each experiment was used for such analyses. Statistical comparisons were made with Origin Pro 2019 (OriginLab Corporation, Northampton, MA) using either ANOVA or unpaired t-test as specified in figure legends. p<0.05 is considered to be statistically significant. The sample size (n) equals the number of cells or membrane patches analyzed. All values are shown as mean ± SE and data graphing was done with Origin Pro 2019.